fix: avoid history pollution in addLinks by bulk-adding edges

Changed addLinks to collect all valid edges first and add them in a single
set() call with one history push, instead of calling addEdge for each link
which would push to history for every edge individually.

Fixes: Sentry bug prediction about undo/redo history pollution

.github/workflows/claude-ci-failure-auto-fix.yml

@@ -40,6 +40,48 @@ jobs:
           git checkout -b "$BRANCH_NAME"
           echo "branch_name=$BRANCH_NAME" >> $GITHUB_OUTPUT
 
+      # Backend Python/Poetry setup (so Claude can run linting/tests)
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Set up Python dependency cache
+        uses: actions/cache@v5
+        with:
+          path: ~/.cache/pypoetry
+          key: poetry-${{ runner.os }}-${{ hashFiles('autogpt_platform/backend/poetry.lock') }}
+
+      - name: Install Poetry
+        run: |
+          cd autogpt_platform/backend
+          HEAD_POETRY_VERSION=$(python3 ../../.github/workflows/scripts/get_package_version_from_lockfile.py poetry)
+          curl -sSL https://install.python-poetry.org | POETRY_VERSION=$HEAD_POETRY_VERSION python3 -
+          echo "$HOME/.local/bin" >> $GITHUB_PATH
+
+      - name: Install Python dependencies
+        working-directory: autogpt_platform/backend
+        run: poetry install
+
+      - name: Generate Prisma Client
+        working-directory: autogpt_platform/backend
+        run: poetry run prisma generate && poetry run gen-prisma-stub
+
+      # Frontend Node.js/pnpm setup (so Claude can run linting/tests)
+      - name: Enable corepack
+        run: corepack enable
+
+      - name: Set up Node.js
+        uses: actions/setup-node@v6
+        with:
+          node-version: "22"
+          cache: "pnpm"
+          cache-dependency-path: autogpt_platform/frontend/pnpm-lock.yaml
+
+      - name: Install JavaScript dependencies
+        working-directory: autogpt_platform/frontend
+        run: pnpm install --frozen-lockfile
+
       - name: Get CI failure details
         id: failure_details
         uses: actions/github-script@v8

.github/workflows/claude-dependabot.yml

@@ -77,27 +77,15 @@ jobs:
         run: poetry run prisma generate && poetry run gen-prisma-stub
 
       # Frontend Node.js/pnpm setup (mirrors platform-frontend-ci.yml)
+      - name: Enable corepack
+        run: corepack enable
+
       - name: Set up Node.js
         uses: actions/setup-node@v6
         with:
           node-version: "22"
-
-      - name: Enable corepack
-        run: corepack enable
-
-      - name: Set pnpm store directory
-        run: |
-          pnpm config set store-dir ~/.pnpm-store
-          echo "PNPM_HOME=$HOME/.pnpm-store" >> $GITHUB_ENV
-
-      - name: Cache frontend dependencies
-        uses: actions/cache@v5
-        with:
-          path: ~/.pnpm-store
-          key: ${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml', 'autogpt_platform/frontend/package.json') }}
-          restore-keys: |
-            ${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml') }}
-            ${{ runner.os }}-pnpm-
+          cache: "pnpm"
+          cache-dependency-path: autogpt_platform/frontend/pnpm-lock.yaml
 
       - name: Install JavaScript dependencies
         working-directory: autogpt_platform/frontend

.github/workflows/claude.yml

@@ -93,27 +93,15 @@ jobs:
         run: poetry run prisma generate && poetry run gen-prisma-stub
 
       # Frontend Node.js/pnpm setup (mirrors platform-frontend-ci.yml)
+      - name: Enable corepack
+        run: corepack enable
+
       - name: Set up Node.js
         uses: actions/setup-node@v6
         with:
           node-version: "22"
-
-      - name: Enable corepack
-        run: corepack enable
-
-      - name: Set pnpm store directory
-        run: |
-          pnpm config set store-dir ~/.pnpm-store
-          echo "PNPM_HOME=$HOME/.pnpm-store" >> $GITHUB_ENV
-
-      - name: Cache frontend dependencies
-        uses: actions/cache@v5
-        with:
-          path: ~/.pnpm-store
-          key: ${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml', 'autogpt_platform/frontend/package.json') }}
-          restore-keys: |
-            ${{ runner.os }}-pnpm-${{ hashFiles('autogpt_platform/frontend/pnpm-lock.yaml') }}
-            ${{ runner.os }}-pnpm-
+          cache: "pnpm"
+          cache-dependency-path: autogpt_platform/frontend/pnpm-lock.yaml
 
       - name: Install JavaScript dependencies
         working-directory: autogpt_platform/frontend

.github/workflows/codeql.yml

@@ -62,7 +62,7 @@ jobs:
 
     # Initializes the CodeQL tools for scanning.
     - name: Initialize CodeQL
-      uses: github/codeql-action/init@v3
+      uses: github/codeql-action/init@v4
       with:
         languages: ${{ matrix.language }}
         build-mode: ${{ matrix.build-mode }}
@@ -93,6 +93,6 @@ jobs:
         exit 1
 
     - name: Perform CodeQL Analysis
-      uses: github/codeql-action/analyze@v3
+      uses: github/codeql-action/analyze@v4
       with:
         category: "/language:${{matrix.language}}"

.github/workflows/docs-claude-review.yml

@@ -7,6 +7,10 @@ on:
       - "docs/integrations/**"
       - "autogpt_platform/backend/backend/blocks/**"
 
+concurrency:
+  group: claude-docs-review-${{ github.event.pull_request.number }}
+  cancel-in-progress: true
+
 jobs:
   claude-review:
     # Only run for PRs from members/collaborators
@@ -91,5 +95,35 @@ jobs:
             3. Read corresponding documentation files to verify accuracy
             4. Provide your feedback as a PR comment
 
+            ## IMPORTANT: Comment Marker
+            Start your PR comment with exactly this HTML comment marker on its own line:
+            <!-- CLAUDE_DOCS_REVIEW -->
+
+            This marker is used to identify and replace your comment on subsequent runs.
+
             Be constructive and specific. If everything looks good, say so!
             If there are issues, explain what's wrong and suggest how to fix it.
+
+      - name: Delete old Claude review comments
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          # Get all comment IDs with our marker, sorted by creation date (oldest first)
+          COMMENT_IDS=$(gh api \
+            repos/${{ github.repository }}/issues/${{ github.event.pull_request.number }}/comments \
+            --jq '[.[] | select(.body | contains("<!-- CLAUDE_DOCS_REVIEW -->"))] | sort_by(.created_at) | .[].id')
+
+          # Count comments
+          COMMENT_COUNT=$(echo "$COMMENT_IDS" | grep -c . || true)
+
+          if [ "$COMMENT_COUNT" -gt 1 ]; then
+            # Delete all but the last (newest) comment
+            echo "$COMMENT_IDS" | head -n -1 | while read -r COMMENT_ID; do
+              if [ -n "$COMMENT_ID" ]; then
+                echo "Deleting old review comment: $COMMENT_ID"
+                gh api -X DELETE repos/${{ github.repository }}/issues/comments/$COMMENT_ID
+              fi
+            done
+          else
+            echo "No old review comments to clean up"
+          fi

.github/workflows/pr-overlap-check.yml

@@ -0,0 +1,39 @@
+name: PR Overlap Detection
+
+on:
+  pull_request:
+    types: [opened, synchronize, reopened]
+    branches:
+      - dev
+      - master
+
+permissions:
+  contents: read
+  pull-requests: write
+
+jobs:
+  check-overlaps:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0  # Need full history for merge testing
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+
+      - name: Configure git
+        run: |
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+          git config user.name "github-actions[bot]"
+
+      - name: Run overlap detection
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        # Always succeed - this check informs contributors, it shouldn't block merging
+        continue-on-error: true
+        run: |
+          python .github/scripts/detect_overlaps.py ${{ github.event.pull_request.number }}

autogpt_platform/backend/.env.default

@@ -104,6 +104,12 @@ TWITTER_CLIENT_SECRET=
 # Make a new workspace for your OAuth APP -- trust me
 # https://linear.app/settings/api/applications/new
 # Callback URL: http://localhost:3000/auth/integrations/oauth_callback
+LINEAR_API_KEY=
+# Linear project and team IDs for the feature request tracker.
+# Find these in your Linear workspace URL: linear.app/<workspace>/project/<project-id>
+# and in team settings. Used by the chat copilot to file and search feature requests.
+LINEAR_FEATURE_REQUEST_PROJECT_ID=
+LINEAR_FEATURE_REQUEST_TEAM_ID=
 LINEAR_CLIENT_ID=
 LINEAR_CLIENT_SECRET=
 

autogpt_platform/backend/Dockerfile

@@ -66,13 +66,19 @@ ENV POETRY_HOME=/opt/poetry \
     DEBIAN_FRONTEND=noninteractive
 ENV PATH=/opt/poetry/bin:$PATH
 
-# Install Python, FFmpeg, and ImageMagick (required for video processing blocks)
+# Install Python, FFmpeg, ImageMagick, and CLI tools for agent use.
+# bubblewrap provides OS-level sandbox (whitelist-only FS + no network)
+# for the bash_exec MCP tool.
 # Using --no-install-recommends saves ~650MB by skipping unnecessary deps like llvm, mesa, etc.
 RUN apt-get update && apt-get install -y --no-install-recommends \
     python3.13 \
     python3-pip \
     ffmpeg \
     imagemagick \
+    jq \
+    ripgrep \
+    tree \
+    bubblewrap \
     && rm -rf /var/lib/apt/lists/*
 
 COPY --from=builder /usr/local/lib/python3* /usr/local/lib/python3*

autogpt_platform/backend/backend/api/conn_manager.py

@@ -122,24 +122,6 @@ class ConnectionManager:
 
         return len(connections)
 
-    async def broadcast_to_all(self, *, method: WSMethod, data: dict) -> int:
-        """Broadcast a message to all active websocket connections."""
-        message = WSMessage(
-            method=method,
-            data=data,
-        ).model_dump_json()
-
-        connections = tuple(self.active_connections)
-        if not connections:
-            return 0
-
-        await asyncio.gather(
-            *(connection.send_text(message) for connection in connections),
-            return_exceptions=True,
-        )
-
-        return len(connections)
-
     async def _subscribe(self, channel_key: str, websocket: WebSocket) -> str:
         if channel_key not in self.subscriptions:
             self.subscriptions[channel_key] = set()

autogpt_platform/backend/backend/api/features/admin/execution_analytics_routes.py

@@ -176,64 +176,30 @@ async def get_execution_analytics_config(
         # Return with provider prefix for clarity
         return f"{provider_name}: {model_name}"
 
-    # Get all models from the registry (dynamic, not hardcoded enum)
-    from backend.data import llm_registry
-    from backend.server.v2.llm import db as llm_db
-
-    # Get the recommended model from the database (configurable via admin UI)
-    recommended_model_slug = await llm_db.get_recommended_model_slug()
-
-    # Build the available models list
-    first_enabled_slug = None
-    for registry_model in llm_registry.iter_dynamic_models():
-        # Only include enabled models in the list
-        if not registry_model.is_enabled:
-            continue
-
-        # Track first enabled model as fallback
-        if first_enabled_slug is None:
-            first_enabled_slug = registry_model.slug
-
-        model = LlmModel(registry_model.slug)
+    # Include all LlmModel values (no more filtering by hardcoded list)
+    recommended_model = LlmModel.GPT4O_MINI.value
+    for model in LlmModel:
         label = generate_model_label(model)
         # Add "(Recommended)" suffix to the recommended model
-        if registry_model.slug == recommended_model_slug:
+        if model.value == recommended_model:
             label += " (Recommended)"
 
         available_models.append(
             ModelInfo(
-                value=registry_model.slug,
+                value=model.value,
                 label=label,
-                provider=registry_model.metadata.provider,
+                provider=model.provider,
             )
         )
 
     # Sort models by provider and name for better UX
     available_models.sort(key=lambda x: (x.provider, x.label))
 
-    # Handle case where no models are available
-    if not available_models:
-        logger.warning(
-            "No enabled LLM models found in registry. "
-            "Ensure models are configured and enabled in the LLM Registry."
-        )
-        # Provide a placeholder entry so admins see meaningful feedback
-        available_models.append(
-            ModelInfo(
-                value="",
-                label="No models available - configure in LLM Registry",
-                provider="none",
-            )
-        )
-
-    # Use the DB recommended model, or fallback to first enabled model
-    final_recommended = recommended_model_slug or first_enabled_slug or ""
-
     return ExecutionAnalyticsConfig(
         available_models=available_models,
         default_system_prompt=DEFAULT_SYSTEM_PROMPT,
         default_user_prompt=DEFAULT_USER_PROMPT,
-        recommended_model=final_recommended,
+        recommended_model=recommended_model,
     )
 
 

autogpt_platform/backend/backend/api/features/admin/llm_routes.py

@@ -1,593 +0,0 @@
-import logging
-
-import autogpt_libs.auth
-import fastapi
-
-from backend.data import llm_registry
-from backend.data.block_cost_config import refresh_llm_costs
-from backend.server.v2.llm import db as llm_db
-from backend.server.v2.llm import model as llm_model
-
-logger = logging.getLogger(__name__)
-
-router = fastapi.APIRouter(
-    tags=["llm", "admin"],
-    dependencies=[fastapi.Security(autogpt_libs.auth.requires_admin_user)],
-)
-
-
-async def _refresh_runtime_state() -> None:
-    """Refresh the LLM registry and clear all related caches to ensure real-time updates."""
-    logger.info("Refreshing LLM registry runtime state...")
-    try:
-        # Refresh registry from database
-        await llm_registry.refresh_llm_registry()
-        await refresh_llm_costs()
-
-        # Clear block schema caches so they're regenerated with updated model options
-        from backend.blocks._base import BlockSchema
-
-        BlockSchema.clear_all_schema_caches()
-        logger.info("Cleared all block schema caches")
-
-        # Clear the /blocks endpoint cache so frontend gets updated schemas
-        try:
-            from backend.api.features.v1 import _get_cached_blocks
-
-            _get_cached_blocks.cache_clear()
-            logger.info("Cleared /blocks endpoint cache")
-        except Exception as e:
-            logger.warning("Failed to clear /blocks cache: %s", e)
-
-        # Clear the v2 builder caches
-        try:
-            from backend.api.features.builder import db as builder_db
-
-            builder_db._get_all_providers.cache_clear()
-            logger.info("Cleared v2 builder providers cache")
-            builder_db._build_cached_search_results.cache_clear()
-            logger.info("Cleared v2 builder search results cache")
-        except Exception as e:
-            logger.debug("Could not clear v2 builder cache: %s", e)
-
-        # Notify all executor services to refresh their registry cache
-        from backend.data.llm_registry import publish_registry_refresh_notification
-
-        await publish_registry_refresh_notification()
-        logger.info("Published registry refresh notification")
-    except Exception as exc:
-        logger.exception(
-            "LLM runtime state refresh failed; caches may be stale: %s", exc
-        )
-
-
-@router.get(
-    "/providers",
-    summary="List LLM providers",
-    response_model=llm_model.LlmProvidersResponse,
-)
-async def list_llm_providers(include_models: bool = True):
-    providers = await llm_db.list_providers(include_models=include_models)
-    return llm_model.LlmProvidersResponse(providers=providers)
-
-
-@router.post(
-    "/providers",
-    summary="Create LLM provider",
-    response_model=llm_model.LlmProvider,
-)
-async def create_llm_provider(request: llm_model.UpsertLlmProviderRequest):
-    provider = await llm_db.upsert_provider(request=request)
-    await _refresh_runtime_state()
-    return provider
-
-
-@router.patch(
-    "/providers/{provider_id}",
-    summary="Update LLM provider",
-    response_model=llm_model.LlmProvider,
-)
-async def update_llm_provider(
-    provider_id: str,
-    request: llm_model.UpsertLlmProviderRequest,
-):
-    provider = await llm_db.upsert_provider(request=request, provider_id=provider_id)
-    await _refresh_runtime_state()
-    return provider
-
-
-@router.delete(
-    "/providers/{provider_id}",
-    summary="Delete LLM provider",
-    response_model=dict,
-)
-async def delete_llm_provider(provider_id: str):
-    """
-    Delete an LLM provider.
-
-    A provider can only be deleted if it has no associated models.
-    Delete all models from the provider first before deleting the provider.
-    """
-    try:
-        await llm_db.delete_provider(provider_id)
-        await _refresh_runtime_state()
-        logger.info("Deleted LLM provider '%s'", provider_id)
-        return {"success": True, "message": "Provider deleted successfully"}
-    except ValueError as e:
-        logger.warning("Failed to delete provider '%s': %s", provider_id, e)
-        raise fastapi.HTTPException(status_code=400, detail=str(e))
-    except Exception as e:
-        logger.exception("Failed to delete provider '%s': %s", provider_id, e)
-        raise fastapi.HTTPException(status_code=500, detail=str(e))
-
-
-@router.get(
-    "/models",
-    summary="List LLM models",
-    response_model=llm_model.LlmModelsResponse,
-)
-async def list_llm_models(
-    provider_id: str | None = fastapi.Query(default=None),
-    page: int = fastapi.Query(default=1, ge=1, description="Page number (1-indexed)"),
-    page_size: int = fastapi.Query(
-        default=50, ge=1, le=100, description="Number of models per page"
-    ),
-):
-    return await llm_db.list_models(
-        provider_id=provider_id, page=page, page_size=page_size
-    )
-
-
-@router.post(
-    "/models",
-    summary="Create LLM model",
-    response_model=llm_model.LlmModel,
-)
-async def create_llm_model(request: llm_model.CreateLlmModelRequest):
-    model = await llm_db.create_model(request=request)
-    await _refresh_runtime_state()
-    return model
-
-
-@router.patch(
-    "/models/{model_id}",
-    summary="Update LLM model",
-    response_model=llm_model.LlmModel,
-)
-async def update_llm_model(
-    model_id: str,
-    request: llm_model.UpdateLlmModelRequest,
-):
-    model = await llm_db.update_model(model_id=model_id, request=request)
-    await _refresh_runtime_state()
-    return model
-
-
-@router.patch(
-    "/models/{model_id}/toggle",
-    summary="Toggle LLM model availability",
-    response_model=llm_model.ToggleLlmModelResponse,
-)
-async def toggle_llm_model(
-    model_id: str,
-    request: llm_model.ToggleLlmModelRequest,
-):
-    """
-    Toggle a model's enabled status, optionally migrating workflows when disabling.
-
-    If disabling a model and `migrate_to_slug` is provided, all workflows using
-    this model will be migrated to the specified replacement model before disabling.
-    A migration record is created which can be reverted later using the revert endpoint.
-
-    Optional fields:
-    - `migration_reason`: Reason for the migration (e.g., "Provider outage")
-    - `custom_credit_cost`: Custom pricing override for billing during migration
-    """
-    try:
-        result = await llm_db.toggle_model(
-            model_id=model_id,
-            is_enabled=request.is_enabled,
-            migrate_to_slug=request.migrate_to_slug,
-            migration_reason=request.migration_reason,
-            custom_credit_cost=request.custom_credit_cost,
-        )
-        await _refresh_runtime_state()
-        if result.nodes_migrated > 0:
-            logger.info(
-                "Toggled model '%s' to %s and migrated %d nodes to '%s' (migration_id=%s)",
-                result.model.slug,
-                "enabled" if request.is_enabled else "disabled",
-                result.nodes_migrated,
-                result.migrated_to_slug,
-                result.migration_id,
-            )
-        return result
-    except ValueError as exc:
-        logger.warning("Model toggle validation failed: %s", exc)
-        raise fastapi.HTTPException(status_code=400, detail=str(exc)) from exc
-    except Exception as exc:
-        logger.exception("Failed to toggle LLM model %s: %s", model_id, exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to toggle model availability",
-        ) from exc
-
-
-@router.get(
-    "/models/{model_id}/usage",
-    summary="Get model usage count",
-    response_model=llm_model.LlmModelUsageResponse,
-)
-async def get_llm_model_usage(model_id: str):
-    """Get the number of workflow nodes using this model."""
-    try:
-        return await llm_db.get_model_usage(model_id=model_id)
-    except ValueError as exc:
-        raise fastapi.HTTPException(status_code=404, detail=str(exc)) from exc
-    except Exception as exc:
-        logger.exception("Failed to get model usage %s: %s", model_id, exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to get model usage",
-        ) from exc
-
-
-@router.delete(
-    "/models/{model_id}",
-    summary="Delete LLM model and migrate workflows",
-    response_model=llm_model.DeleteLlmModelResponse,
-)
-async def delete_llm_model(
-    model_id: str,
-    replacement_model_slug: str | None = fastapi.Query(
-        default=None,
-        description="Slug of the model to migrate existing workflows to (required only if workflows use this model)",
-    ),
-):
-    """
-    Delete a model and optionally migrate workflows using it to a replacement model.
-
-    If no workflows are using this model, it can be deleted without providing a
-    replacement. If workflows exist, replacement_model_slug is required.
-
-    This endpoint:
-    1. Counts how many workflow nodes use the model being deleted
-    2. If nodes exist, validates the replacement model and migrates them
-    3. Deletes the model record
-    4. Refreshes all caches and notifies executors
-
-    Example: DELETE /api/llm/admin/models/{id}?replacement_model_slug=gpt-4o
-    Example (no usage): DELETE /api/llm/admin/models/{id}
-    """
-    try:
-        result = await llm_db.delete_model(
-            model_id=model_id, replacement_model_slug=replacement_model_slug
-        )
-        await _refresh_runtime_state()
-        logger.info(
-            "Deleted model '%s' and migrated %d nodes to '%s'",
-            result.deleted_model_slug,
-            result.nodes_migrated,
-            result.replacement_model_slug,
-        )
-        return result
-    except ValueError as exc:
-        # Validation errors (model not found, replacement invalid, etc.)
-        logger.warning("Model deletion validation failed: %s", exc)
-        raise fastapi.HTTPException(status_code=400, detail=str(exc)) from exc
-    except Exception as exc:
-        logger.exception("Failed to delete LLM model %s: %s", model_id, exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to delete model and migrate workflows",
-        ) from exc
-
-
-# ============================================================================
-# Migration Management Endpoints
-# ============================================================================
-
-
-@router.get(
-    "/migrations",
-    summary="List model migrations",
-    response_model=llm_model.LlmMigrationsResponse,
-)
-async def list_llm_migrations(
-    include_reverted: bool = fastapi.Query(
-        default=False, description="Include reverted migrations in the list"
-    ),
-):
-    """
-    List all model migrations.
-
-    Migrations are created when disabling a model with the migrate_to_slug option.
-    They can be reverted to restore the original model configuration.
-    """
-    try:
-        migrations = await llm_db.list_migrations(include_reverted=include_reverted)
-        return llm_model.LlmMigrationsResponse(migrations=migrations)
-    except Exception as exc:
-        logger.exception("Failed to list migrations: %s", exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to list migrations",
-        ) from exc
-
-
-@router.get(
-    "/migrations/{migration_id}",
-    summary="Get migration details",
-    response_model=llm_model.LlmModelMigration,
-)
-async def get_llm_migration(migration_id: str):
-    """Get details of a specific migration."""
-    try:
-        migration = await llm_db.get_migration(migration_id)
-        if not migration:
-            raise fastapi.HTTPException(
-                status_code=404, detail=f"Migration '{migration_id}' not found"
-            )
-        return migration
-    except fastapi.HTTPException:
-        raise
-    except Exception as exc:
-        logger.exception("Failed to get migration %s: %s", migration_id, exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to get migration",
-        ) from exc
-
-
-@router.post(
-    "/migrations/{migration_id}/revert",
-    summary="Revert a model migration",
-    response_model=llm_model.RevertMigrationResponse,
-)
-async def revert_llm_migration(
-    migration_id: str,
-    request: llm_model.RevertMigrationRequest | None = None,
-):
-    """
-    Revert a model migration, restoring affected workflows to their original model.
-
-    This only reverts the specific nodes that were part of the migration.
-    The source model must exist for the revert to succeed.
-
-    Options:
-    - `re_enable_source_model`: Whether to re-enable the source model if disabled (default: True)
-
-    Response includes:
-    - `nodes_reverted`: Number of nodes successfully reverted
-    - `nodes_already_changed`: Number of nodes that were modified since migration (not reverted)
-    - `source_model_re_enabled`: Whether the source model was re-enabled
-
-    Requirements:
-    - Migration must not already be reverted
-    - Source model must exist
-    """
-    try:
-        re_enable = request.re_enable_source_model if request else True
-        result = await llm_db.revert_migration(
-            migration_id,
-            re_enable_source_model=re_enable,
-        )
-        await _refresh_runtime_state()
-        logger.info(
-            "Reverted migration '%s': %d nodes restored from '%s' to '%s' "
-            "(%d already changed, source re-enabled=%s)",
-            migration_id,
-            result.nodes_reverted,
-            result.target_model_slug,
-            result.source_model_slug,
-            result.nodes_already_changed,
-            result.source_model_re_enabled,
-        )
-        return result
-    except ValueError as exc:
-        logger.warning("Migration revert validation failed: %s", exc)
-        raise fastapi.HTTPException(status_code=400, detail=str(exc)) from exc
-    except Exception as exc:
-        logger.exception("Failed to revert migration %s: %s", migration_id, exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to revert migration",
-        ) from exc
-
-
-# ============================================================================
-# Creator Management Endpoints
-# ============================================================================
-
-
-@router.get(
-    "/creators",
-    summary="List model creators",
-    response_model=llm_model.LlmCreatorsResponse,
-)
-async def list_llm_creators():
-    """
-    List all model creators.
-
-    Creators are organizations that create/train models (e.g., OpenAI, Meta, Anthropic).
-    This is distinct from providers who host/serve the models (e.g., OpenRouter).
-    """
-    try:
-        creators = await llm_db.list_creators()
-        return llm_model.LlmCreatorsResponse(creators=creators)
-    except Exception as exc:
-        logger.exception("Failed to list creators: %s", exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to list creators",
-        ) from exc
-
-
-@router.get(
-    "/creators/{creator_id}",
-    summary="Get creator details",
-    response_model=llm_model.LlmModelCreator,
-)
-async def get_llm_creator(creator_id: str):
-    """Get details of a specific model creator."""
-    try:
-        creator = await llm_db.get_creator(creator_id)
-        if not creator:
-            raise fastapi.HTTPException(
-                status_code=404, detail=f"Creator '{creator_id}' not found"
-            )
-        return creator
-    except fastapi.HTTPException:
-        raise
-    except Exception as exc:
-        logger.exception("Failed to get creator %s: %s", creator_id, exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to get creator",
-        ) from exc
-
-
-@router.post(
-    "/creators",
-    summary="Create model creator",
-    response_model=llm_model.LlmModelCreator,
-)
-async def create_llm_creator(request: llm_model.UpsertLlmCreatorRequest):
-    """
-    Create a new model creator.
-
-    A creator represents an organization that creates/trains AI models,
-    such as OpenAI, Anthropic, Meta, or Google.
-    """
-    try:
-        creator = await llm_db.upsert_creator(request=request)
-        await _refresh_runtime_state()
-        logger.info("Created model creator '%s' (%s)", creator.display_name, creator.id)
-        return creator
-    except Exception as exc:
-        logger.exception("Failed to create creator: %s", exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to create creator",
-        ) from exc
-
-
-@router.patch(
-    "/creators/{creator_id}",
-    summary="Update model creator",
-    response_model=llm_model.LlmModelCreator,
-)
-async def update_llm_creator(
-    creator_id: str,
-    request: llm_model.UpsertLlmCreatorRequest,
-):
-    """Update an existing model creator."""
-    try:
-        creator = await llm_db.upsert_creator(request=request, creator_id=creator_id)
-        await _refresh_runtime_state()
-        logger.info("Updated model creator '%s' (%s)", creator.display_name, creator_id)
-        return creator
-    except Exception as exc:
-        logger.exception("Failed to update creator %s: %s", creator_id, exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to update creator",
-        ) from exc
-
-
-@router.delete(
-    "/creators/{creator_id}",
-    summary="Delete model creator",
-    response_model=dict,
-)
-async def delete_llm_creator(creator_id: str):
-    """
-    Delete a model creator.
-
-    This will remove the creator association from all models that reference it
-    (sets creatorId to NULL), but will not delete the models themselves.
-    """
-    try:
-        await llm_db.delete_creator(creator_id)
-        await _refresh_runtime_state()
-        logger.info("Deleted model creator '%s'", creator_id)
-        return {"success": True, "message": f"Creator '{creator_id}' deleted"}
-    except ValueError as exc:
-        logger.warning("Creator deletion validation failed: %s", exc)
-        raise fastapi.HTTPException(status_code=404, detail=str(exc)) from exc
-    except Exception as exc:
-        logger.exception("Failed to delete creator %s: %s", creator_id, exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to delete creator",
-        ) from exc
-
-
-# ============================================================================
-# Recommended Model Endpoints
-# ============================================================================
-
-
-@router.get(
-    "/recommended-model",
-    summary="Get recommended model",
-    response_model=llm_model.RecommendedModelResponse,
-)
-async def get_recommended_model():
-    """
-    Get the currently recommended LLM model.
-
-    The recommended model is shown to users as the default/suggested option
-    in model selection dropdowns.
-    """
-    try:
-        model = await llm_db.get_recommended_model()
-        return llm_model.RecommendedModelResponse(
-            model=model,
-            slug=model.slug if model else None,
-        )
-    except Exception as exc:
-        logger.exception("Failed to get recommended model: %s", exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to get recommended model",
-        ) from exc
-
-
-@router.post(
-    "/recommended-model",
-    summary="Set recommended model",
-    response_model=llm_model.SetRecommendedModelResponse,
-)
-async def set_recommended_model(request: llm_model.SetRecommendedModelRequest):
-    """
-    Set a model as the recommended model.
-
-    This clears the recommended flag from any other model and sets it on
-    the specified model. The model must be enabled to be set as recommended.
-
-    The recommended model is displayed to users as the default/suggested
-    option in model selection dropdowns throughout the platform.
-    """
-    try:
-        model, previous_slug = await llm_db.set_recommended_model(request.model_id)
-        await _refresh_runtime_state()
-        logger.info(
-            "Set recommended model to '%s' (previous: %s)",
-            model.slug,
-            previous_slug or "none",
-        )
-        return llm_model.SetRecommendedModelResponse(
-            model=model,
-            previous_recommended_slug=previous_slug,
-            message=f"Model '{model.display_name}' is now the recommended model",
-        )
-    except ValueError as exc:
-        logger.warning("Set recommended model validation failed: %s", exc)
-        raise fastapi.HTTPException(status_code=400, detail=str(exc)) from exc
-    except Exception as exc:
-        logger.exception("Failed to set recommended model: %s", exc)
-        raise fastapi.HTTPException(
-            status_code=500,
-            detail="Failed to set recommended model",
-        ) from exc

autogpt_platform/backend/backend/api/features/admin/llm_routes_test.py

@@ -1,491 +0,0 @@
-import json
-from unittest.mock import AsyncMock
-
-import fastapi
-import fastapi.testclient
-import pytest
-import pytest_mock
-from autogpt_libs.auth.jwt_utils import get_jwt_payload
-from pytest_snapshot.plugin import Snapshot
-
-import backend.api.features.admin.llm_routes as llm_routes
-from backend.server.v2.llm import model as llm_model
-from backend.util.models import Pagination
-
-app = fastapi.FastAPI()
-app.include_router(llm_routes.router, prefix="/admin/llm")
-
-client = fastapi.testclient.TestClient(app)
-
-
-@pytest.fixture(autouse=True)
-def setup_app_admin_auth(mock_jwt_admin):
-    """Setup admin auth overrides for all tests in this module"""
-    app.dependency_overrides[get_jwt_payload] = mock_jwt_admin["get_jwt_payload"]
-    yield
-    app.dependency_overrides.clear()
-
-
-def test_list_llm_providers_success(
-    mocker: pytest_mock.MockFixture,
-    configured_snapshot: Snapshot,
-) -> None:
-    """Test successful listing of LLM providers"""
-    # Mock the database function
-    mock_providers = [
-        {
-            "id": "provider-1",
-            "name": "openai",
-            "display_name": "OpenAI",
-            "description": "OpenAI LLM provider",
-            "supports_tools": True,
-            "supports_json_output": True,
-            "supports_reasoning": False,
-            "supports_parallel_tool": True,
-            "metadata": {},
-            "models": [],
-        },
-        {
-            "id": "provider-2",
-            "name": "anthropic",
-            "display_name": "Anthropic",
-            "description": "Anthropic LLM provider",
-            "supports_tools": True,
-            "supports_json_output": True,
-            "supports_reasoning": False,
-            "supports_parallel_tool": True,
-            "metadata": {},
-            "models": [],
-        },
-    ]
-
-    mocker.patch(
-        "backend.api.features.admin.llm_routes.llm_db.list_providers",
-        new=AsyncMock(return_value=mock_providers),
-    )
-
-    response = client.get("/admin/llm/providers")
-
-    assert response.status_code == 200
-    response_data = response.json()
-    assert len(response_data["providers"]) == 2
-    assert response_data["providers"][0]["name"] == "openai"
-
-    # Snapshot test the response (must be string)
-    configured_snapshot.assert_match(
-        json.dumps(response_data, indent=2, sort_keys=True),
-        "list_llm_providers_success.json",
-    )
-
-
-def test_list_llm_models_success(
-    mocker: pytest_mock.MockFixture,
-    configured_snapshot: Snapshot,
-) -> None:
-    """Test successful listing of LLM models with pagination"""
-    # Mock the database function - now returns LlmModelsResponse
-    mock_model = llm_model.LlmModel(
-        id="model-1",
-        slug="gpt-4o",
-        display_name="GPT-4o",
-        description="GPT-4 Optimized",
-        provider_id="provider-1",
-        context_window=128000,
-        max_output_tokens=16384,
-        is_enabled=True,
-        capabilities={},
-        metadata={},
-        costs=[
-            llm_model.LlmModelCost(
-                id="cost-1",
-                credit_cost=10,
-                credential_provider="openai",
-                metadata={},
-            )
-        ],
-    )
-
-    mock_response = llm_model.LlmModelsResponse(
-        models=[mock_model],
-        pagination=Pagination(
-            total_items=1,
-            total_pages=1,
-            current_page=1,
-            page_size=50,
-        ),
-    )
-
-    mocker.patch(
-        "backend.api.features.admin.llm_routes.llm_db.list_models",
-        new=AsyncMock(return_value=mock_response),
-    )
-
-    response = client.get("/admin/llm/models")
-
-    assert response.status_code == 200
-    response_data = response.json()
-    assert len(response_data["models"]) == 1
-    assert response_data["models"][0]["slug"] == "gpt-4o"
-    assert response_data["pagination"]["total_items"] == 1
-    assert response_data["pagination"]["page_size"] == 50
-
-    # Snapshot test the response (must be string)
-    configured_snapshot.assert_match(
-        json.dumps(response_data, indent=2, sort_keys=True),
-        "list_llm_models_success.json",
-    )
-
-
-def test_create_llm_provider_success(
-    mocker: pytest_mock.MockFixture,
-    configured_snapshot: Snapshot,
-) -> None:
-    """Test successful creation of LLM provider"""
-    mock_provider = {
-        "id": "new-provider-id",
-        "name": "groq",
-        "display_name": "Groq",
-        "description": "Groq LLM provider",
-        "supports_tools": True,
-        "supports_json_output": True,
-        "supports_reasoning": False,
-        "supports_parallel_tool": False,
-        "metadata": {},
-    }
-
-    mocker.patch(
-        "backend.api.features.admin.llm_routes.llm_db.upsert_provider",
-        new=AsyncMock(return_value=mock_provider),
-    )
-
-    mock_refresh = mocker.patch(
-        "backend.api.features.admin.llm_routes._refresh_runtime_state",
-        new=AsyncMock(),
-    )
-
-    request_data = {
-        "name": "groq",
-        "display_name": "Groq",
-        "description": "Groq LLM provider",
-        "supports_tools": True,
-        "supports_json_output": True,
-        "supports_reasoning": False,
-        "supports_parallel_tool": False,
-        "metadata": {},
-    }
-
-    response = client.post("/admin/llm/providers", json=request_data)
-
-    assert response.status_code == 200
-    response_data = response.json()
-    assert response_data["name"] == "groq"
-    assert response_data["display_name"] == "Groq"
-
-    # Verify refresh was called
-    mock_refresh.assert_called_once()
-
-    # Snapshot test the response (must be string)
-    configured_snapshot.assert_match(
-        json.dumps(response_data, indent=2, sort_keys=True),
-        "create_llm_provider_success.json",
-    )
-
-
-def test_create_llm_model_success(
-    mocker: pytest_mock.MockFixture,
-    configured_snapshot: Snapshot,
-) -> None:
-    """Test successful creation of LLM model"""
-    mock_model = {
-        "id": "new-model-id",
-        "slug": "gpt-4.1-mini",
-        "display_name": "GPT-4.1 Mini",
-        "description": "Latest GPT-4.1 Mini model",
-        "provider_id": "provider-1",
-        "context_window": 128000,
-        "max_output_tokens": 16384,
-        "is_enabled": True,
-        "capabilities": {},
-        "metadata": {},
-        "costs": [
-            {
-                "id": "cost-id",
-                "credit_cost": 5,
-                "credential_provider": "openai",
-                "metadata": {},
-            }
-        ],
-    }
-
-    mocker.patch(
-        "backend.api.features.admin.llm_routes.llm_db.create_model",
-        new=AsyncMock(return_value=mock_model),
-    )
-
-    mock_refresh = mocker.patch(
-        "backend.api.features.admin.llm_routes._refresh_runtime_state",
-        new=AsyncMock(),
-    )
-
-    request_data = {
-        "slug": "gpt-4.1-mini",
-        "display_name": "GPT-4.1 Mini",
-        "description": "Latest GPT-4.1 Mini model",
-        "provider_id": "provider-1",
-        "context_window": 128000,
-        "max_output_tokens": 16384,
-        "is_enabled": True,
-        "capabilities": {},
-        "metadata": {},
-        "costs": [
-            {
-                "credit_cost": 5,
-                "credential_provider": "openai",
-                "metadata": {},
-            }
-        ],
-    }
-
-    response = client.post("/admin/llm/models", json=request_data)
-
-    assert response.status_code == 200
-    response_data = response.json()
-    assert response_data["slug"] == "gpt-4.1-mini"
-    assert response_data["is_enabled"] is True
-
-    # Verify refresh was called
-    mock_refresh.assert_called_once()
-
-    # Snapshot test the response (must be string)
-    configured_snapshot.assert_match(
-        json.dumps(response_data, indent=2, sort_keys=True),
-        "create_llm_model_success.json",
-    )
-
-
-def test_update_llm_model_success(
-    mocker: pytest_mock.MockFixture,
-    configured_snapshot: Snapshot,
-) -> None:
-    """Test successful update of LLM model"""
-    mock_model = {
-        "id": "model-1",
-        "slug": "gpt-4o",
-        "display_name": "GPT-4o Updated",
-        "description": "Updated description",
-        "provider_id": "provider-1",
-        "context_window": 256000,
-        "max_output_tokens": 32768,
-        "is_enabled": True,
-        "capabilities": {},
-        "metadata": {},
-        "costs": [
-            {
-                "id": "cost-1",
-                "credit_cost": 15,
-                "credential_provider": "openai",
-                "metadata": {},
-            }
-        ],
-    }
-
-    mocker.patch(
-        "backend.api.features.admin.llm_routes.llm_db.update_model",
-        new=AsyncMock(return_value=mock_model),
-    )
-
-    mock_refresh = mocker.patch(
-        "backend.api.features.admin.llm_routes._refresh_runtime_state",
-        new=AsyncMock(),
-    )
-
-    request_data = {
-        "display_name": "GPT-4o Updated",
-        "description": "Updated description",
-        "context_window": 256000,
-        "max_output_tokens": 32768,
-    }
-
-    response = client.patch("/admin/llm/models/model-1", json=request_data)
-
-    assert response.status_code == 200
-    response_data = response.json()
-    assert response_data["display_name"] == "GPT-4o Updated"
-    assert response_data["context_window"] == 256000
-
-    # Verify refresh was called
-    mock_refresh.assert_called_once()
-
-    # Snapshot test the response (must be string)
-    configured_snapshot.assert_match(
-        json.dumps(response_data, indent=2, sort_keys=True),
-        "update_llm_model_success.json",
-    )
-
-
-def test_toggle_llm_model_success(
-    mocker: pytest_mock.MockFixture,
-    configured_snapshot: Snapshot,
-) -> None:
-    """Test successful toggling of LLM model enabled status"""
-    # Create a proper mock model object
-    mock_model = llm_model.LlmModel(
-        id="model-1",
-        slug="gpt-4o",
-        display_name="GPT-4o",
-        description="GPT-4 Optimized",
-        provider_id="provider-1",
-        context_window=128000,
-        max_output_tokens=16384,
-        is_enabled=False,
-        capabilities={},
-        metadata={},
-        costs=[],
-    )
-
-    # Create a proper ToggleLlmModelResponse
-    mock_response = llm_model.ToggleLlmModelResponse(
-        model=mock_model,
-        nodes_migrated=0,
-        migrated_to_slug=None,
-        migration_id=None,
-    )
-
-    mocker.patch(
-        "backend.api.features.admin.llm_routes.llm_db.toggle_model",
-        new=AsyncMock(return_value=mock_response),
-    )
-
-    mock_refresh = mocker.patch(
-        "backend.api.features.admin.llm_routes._refresh_runtime_state",
-        new=AsyncMock(),
-    )
-
-    request_data = {"is_enabled": False}
-
-    response = client.patch("/admin/llm/models/model-1/toggle", json=request_data)
-
-    assert response.status_code == 200
-    response_data = response.json()
-    assert response_data["model"]["is_enabled"] is False
-
-    # Verify refresh was called
-    mock_refresh.assert_called_once()
-
-    # Snapshot test the response (must be string)
-    configured_snapshot.assert_match(
-        json.dumps(response_data, indent=2, sort_keys=True),
-        "toggle_llm_model_success.json",
-    )
-
-
-def test_delete_llm_model_success(
-    mocker: pytest_mock.MockFixture,
-    configured_snapshot: Snapshot,
-) -> None:
-    """Test successful deletion of LLM model with migration"""
-    # Create a proper DeleteLlmModelResponse
-    mock_response = llm_model.DeleteLlmModelResponse(
-        deleted_model_slug="gpt-3.5-turbo",
-        deleted_model_display_name="GPT-3.5 Turbo",
-        replacement_model_slug="gpt-4o-mini",
-        nodes_migrated=42,
-        message="Successfully deleted model 'GPT-3.5 Turbo' (gpt-3.5-turbo) "
-        "and migrated 42 workflow node(s) to 'gpt-4o-mini'.",
-    )
-
-    mocker.patch(
-        "backend.api.features.admin.llm_routes.llm_db.delete_model",
-        new=AsyncMock(return_value=mock_response),
-    )
-
-    mock_refresh = mocker.patch(
-        "backend.api.features.admin.llm_routes._refresh_runtime_state",
-        new=AsyncMock(),
-    )
-
-    response = client.delete(
-        "/admin/llm/models/model-1?replacement_model_slug=gpt-4o-mini"
-    )
-
-    assert response.status_code == 200
-    response_data = response.json()
-    assert response_data["deleted_model_slug"] == "gpt-3.5-turbo"
-    assert response_data["nodes_migrated"] == 42
-    assert response_data["replacement_model_slug"] == "gpt-4o-mini"
-
-    # Verify refresh was called
-    mock_refresh.assert_called_once()
-
-    # Snapshot test the response (must be string)
-    configured_snapshot.assert_match(
-        json.dumps(response_data, indent=2, sort_keys=True),
-        "delete_llm_model_success.json",
-    )
-
-
-def test_delete_llm_model_validation_error(
-    mocker: pytest_mock.MockFixture,
-) -> None:
-    """Test deletion fails with proper error when validation fails"""
-    mocker.patch(
-        "backend.api.features.admin.llm_routes.llm_db.delete_model",
-        new=AsyncMock(side_effect=ValueError("Replacement model 'invalid' not found")),
-    )
-
-    response = client.delete("/admin/llm/models/model-1?replacement_model_slug=invalid")
-
-    assert response.status_code == 400
-    assert "Replacement model 'invalid' not found" in response.json()["detail"]
-
-
-def test_delete_llm_model_no_replacement_with_usage(
-    mocker: pytest_mock.MockFixture,
-) -> None:
-    """Test deletion fails when nodes exist but no replacement is provided"""
-    mocker.patch(
-        "backend.api.features.admin.llm_routes.llm_db.delete_model",
-        new=AsyncMock(
-            side_effect=ValueError(
-                "Cannot delete model 'test-model': 5 workflow node(s) are using it. "
-                "Please provide a replacement_model_slug to migrate them."
-            )
-        ),
-    )
-
-    response = client.delete("/admin/llm/models/model-1")
-
-    assert response.status_code == 400
-    assert "workflow node(s) are using it" in response.json()["detail"]
-
-
-def test_delete_llm_model_no_replacement_no_usage(
-    mocker: pytest_mock.MockFixture,
-) -> None:
-    """Test deletion succeeds when no nodes use the model and no replacement is provided"""
-    mock_response = llm_model.DeleteLlmModelResponse(
-        deleted_model_slug="unused-model",
-        deleted_model_display_name="Unused Model",
-        replacement_model_slug=None,
-        nodes_migrated=0,
-        message="Successfully deleted model 'Unused Model' (unused-model). No workflows were using this model.",
-    )
-
-    mocker.patch(
-        "backend.api.features.admin.llm_routes.llm_db.delete_model",
-        new=AsyncMock(return_value=mock_response),
-    )
-
-    mock_refresh = mocker.patch(
-        "backend.api.features.admin.llm_routes._refresh_runtime_state",
-        new=AsyncMock(),
-    )
-
-    response = client.delete("/admin/llm/models/model-1")
-
-    assert response.status_code == 200
-    response_data = response.json()
-    assert response_data["deleted_model_slug"] == "unused-model"
-    assert response_data["nodes_migrated"] == 0
-    assert response_data["replacement_model_slug"] is None
-    mock_refresh.assert_called_once()

autogpt_platform/backend/backend/api/features/builder/db.py

@@ -20,7 +20,6 @@ from backend.blocks._base import (
 )
 from backend.blocks.llm import LlmModel
 from backend.data.db import query_raw_with_schema
-from backend.data.llm_registry import get_all_model_slugs_for_validation
 from backend.integrations.providers import ProviderName
 from backend.util.cache import cached
 from backend.util.models import Pagination
@@ -37,14 +36,7 @@ from .model import (
 )
 
 logger = logging.getLogger(__name__)
-
-
-def _get_llm_models() -> list[str]:
-    """Get LLM model names for search matching from the registry."""
-    return [
-        slug.lower().replace("-", " ") for slug in get_all_model_slugs_for_validation()
-    ]
-
+llm_models = [name.name.lower().replace("_", " ") for name in LlmModel]
 
 MAX_LIBRARY_AGENT_RESULTS = 100
 MAX_MARKETPLACE_AGENT_RESULTS = 100
@@ -509,10 +501,8 @@ async def _get_static_counts():
 def _matches_llm_model(schema_cls: type[BlockSchema], query: str) -> bool:
     for field in schema_cls.model_fields.values():
         if field.annotation == LlmModel:
-            # Normalize query same as model slugs (lowercase, hyphens to spaces)
-            normalized_model_query = query.lower().replace("-", " ")
-            # Check if query matches any value in llm_models from registry
-            if any(normalized_model_query in name for name in _get_llm_models()):
+            # Check if query matches any value in llm_models
+            if any(query in name for name in llm_models):
                 return True
     return False
 

autogpt_platform/backend/backend/api/features/chat/config.py

@@ -27,12 +27,11 @@ class ChatConfig(BaseSettings):
     session_ttl: int = Field(default=43200, description="Session TTL in seconds")
 
     # Streaming Configuration
-    max_context_messages: int = Field(
-        default=50, ge=1, le=200, description="Maximum context messages"
-    )
-
     stream_timeout: int = Field(default=300, description="Stream timeout in seconds")
-    max_retries: int = Field(default=3, description="Maximum number of retries")
+    max_retries: int = Field(
+        default=3,
+        description="Max retries for fallback path (SDK handles retries internally)",
+    )
     max_agent_runs: int = Field(default=30, description="Maximum number of agent runs")
     max_agent_schedules: int = Field(
         default=30, description="Maximum number of agent schedules"
@@ -93,6 +92,31 @@ class ChatConfig(BaseSettings):
         description="Name of the prompt in Langfuse to fetch",
     )
 
+    # Claude Agent SDK Configuration
+    use_claude_agent_sdk: bool = Field(
+        default=True,
+        description="Use Claude Agent SDK for chat completions",
+    )
+    claude_agent_model: str | None = Field(
+        default=None,
+        description="Model for the Claude Agent SDK path. If None, derives from "
+        "the `model` field by stripping the OpenRouter provider prefix.",
+    )
+    claude_agent_max_buffer_size: int = Field(
+        default=10 * 1024 * 1024,  # 10MB (default SDK is 1MB)
+        description="Max buffer size in bytes for Claude Agent SDK JSON message parsing. "
+        "Increase if tool outputs exceed the limit.",
+    )
+    claude_agent_max_subtasks: int = Field(
+        default=10,
+        description="Max number of sub-agent Tasks the SDK can spawn per session.",
+    )
+    claude_agent_use_resume: bool = Field(
+        default=True,
+        description="Use --resume for multi-turn conversations instead of "
+        "history compression. Falls back to compression when unavailable.",
+    )
+
     # Extended thinking configuration for Claude models
     thinking_enabled: bool = Field(
         default=True,
@@ -138,6 +162,17 @@ class ChatConfig(BaseSettings):
             v = os.getenv("CHAT_INTERNAL_API_KEY")
         return v
 
+    @field_validator("use_claude_agent_sdk", mode="before")
+    @classmethod
+    def get_use_claude_agent_sdk(cls, v):
+        """Get use_claude_agent_sdk from environment if not provided."""
+        # Check environment variable - default to True if not set
+        env_val = os.getenv("CHAT_USE_CLAUDE_AGENT_SDK", "").lower()
+        if env_val:
+            return env_val in ("true", "1", "yes", "on")
+        # Default to True (SDK enabled by default)
+        return True if v is None else v
+
     # Prompt paths for different contexts
     PROMPT_PATHS: dict[str, str] = {
         "default": "prompts/chat_system.md",

autogpt_platform/backend/backend/api/features/chat/model.py

@@ -334,9 +334,8 @@ async def _get_session_from_cache(session_id: str) -> ChatSession | None:
     try:
         session = ChatSession.model_validate_json(raw_session)
         logger.info(
-            f"Loading session {session_id} from cache: "
-            f"message_count={len(session.messages)}, "
-            f"roles={[m.role for m in session.messages]}"
+            f"[CACHE] Loaded session {session_id}: {len(session.messages)} messages, "
+            f"last_roles={[m.role for m in session.messages[-3:]]}"  # Last 3 roles
         )
         return session
     except Exception as e:
@@ -378,11 +377,9 @@ async def _get_session_from_db(session_id: str) -> ChatSession | None:
         return None
 
     messages = prisma_session.Messages
-    logger.info(
-        f"Loading session {session_id} from DB: "
-        f"has_messages={messages is not None}, "
-        f"message_count={len(messages) if messages else 0}, "
-        f"roles={[m.role for m in messages] if messages else []}"
+    logger.debug(
+        f"[DB] Loaded session {session_id}: {len(messages) if messages else 0} messages, "
+        f"roles={[m.role for m in messages[-3:]] if messages else []}"  # Last 3 roles
     )
 
     return ChatSession.from_db(prisma_session, messages)
@@ -433,10 +430,9 @@ async def _save_session_to_db(
                     "function_call": msg.function_call,
                 }
             )
-        logger.info(
-            f"Saving {len(new_messages)} new messages to DB for session {session.session_id}: "
-            f"roles={[m['role'] for m in messages_data]}, "
-            f"start_sequence={existing_message_count}"
+        logger.debug(
+            f"[DB] Saving {len(new_messages)} messages to session {session.session_id}, "
+            f"roles={[m['role'] for m in messages_data]}"
         )
         await chat_db.add_chat_messages_batch(
             session_id=session.session_id,
@@ -476,7 +472,7 @@ async def get_chat_session(
         logger.warning(f"Unexpected cache error for session {session_id}: {e}")
 
     # Fall back to database
-    logger.info(f"Session {session_id} not in cache, checking database")
+    logger.debug(f"Session {session_id} not in cache, checking database")
     session = await _get_session_from_db(session_id)
 
     if session is None:
@@ -493,7 +489,6 @@ async def get_chat_session(
     # Cache the session from DB
     try:
         await _cache_session(session)
-        logger.info(f"Cached session {session_id} from database")
     except Exception as e:
         logger.warning(f"Failed to cache session {session_id}: {e}")
 
@@ -558,6 +553,40 @@ async def upsert_chat_session(
         return session
 
 
+async def append_and_save_message(session_id: str, message: ChatMessage) -> ChatSession:
+    """Atomically append a message to a session and persist it.
+
+    Acquires the session lock, re-fetches the latest session state,
+    appends the message, and saves — preventing message loss when
+    concurrent requests modify the same session.
+    """
+    lock = await _get_session_lock(session_id)
+
+    async with lock:
+        session = await get_chat_session(session_id)
+        if session is None:
+            raise ValueError(f"Session {session_id} not found")
+
+        session.messages.append(message)
+        existing_message_count = await chat_db.get_chat_session_message_count(
+            session_id
+        )
+
+        try:
+            await _save_session_to_db(session, existing_message_count)
+        except Exception as e:
+            raise DatabaseError(
+                f"Failed to persist message to session {session_id}"
+            ) from e
+
+        try:
+            await _cache_session(session)
+        except Exception as e:
+            logger.warning(f"Cache write failed for session {session_id}: {e}")
+
+        return session
+
+
 async def create_chat_session(user_id: str) -> ChatSession:
     """Create a new chat session and persist it.
 
@@ -664,13 +693,19 @@ async def update_session_title(session_id: str, title: str) -> bool:
             logger.warning(f"Session {session_id} not found for title update")
             return False
 
-        # Invalidate cache so next fetch gets updated title
+        # Update title in cache if it exists (instead of invalidating).
+        # This prevents race conditions where cache invalidation causes
+        # the frontend to see stale DB data while streaming is still in progress.
         try:
-            redis_key = _get_session_cache_key(session_id)
-            async_redis = await get_redis_async()
-            await async_redis.delete(redis_key)
+            cached = await _get_session_from_cache(session_id)
+            if cached:
+                cached.title = title
+                await _cache_session(cached)
         except Exception as e:
-            logger.warning(f"Failed to invalidate cache for session {session_id}: {e}")
+            # Not critical - title will be correct on next full cache refresh
+            logger.warning(
+                f"Failed to update title in cache for session {session_id}: {e}"
+            )
 
         return True
     except Exception as e:

autogpt_platform/backend/backend/api/features/chat/routes.py

@@ -1,5 +1,6 @@
 """Chat API routes for chat session management and streaming via SSE."""
 
+import asyncio
 import logging
 import uuid as uuid_module
 from collections.abc import AsyncGenerator
@@ -11,13 +12,22 @@ from fastapi.responses import StreamingResponse
 from pydantic import BaseModel
 
 from backend.util.exceptions import NotFoundError
+from backend.util.feature_flag import Flag, is_feature_enabled
 
 from . import service as chat_service
 from . import stream_registry
 from .completion_handler import process_operation_failure, process_operation_success
 from .config import ChatConfig
-from .model import ChatSession, create_chat_session, get_chat_session, get_user_sessions
-from .response_model import StreamFinish, StreamHeartbeat
+from .model import (
+    ChatMessage,
+    ChatSession,
+    append_and_save_message,
+    create_chat_session,
+    get_chat_session,
+    get_user_sessions,
+)
+from .response_model import StreamError, StreamFinish, StreamHeartbeat, StreamStart
+from .sdk import service as sdk_service
 from .tools.models import (
     AgentDetailsResponse,
     AgentOutputResponse,
@@ -41,6 +51,7 @@ from .tools.models import (
     SetupRequirementsResponse,
     UnderstandingUpdatedResponse,
 )
+from .tracking import track_user_message
 
 config = ChatConfig()
 
@@ -232,6 +243,10 @@ async def get_session(
     active_task, last_message_id = await stream_registry.get_active_task_for_session(
         session_id, user_id
     )
+    logger.info(
+        f"[GET_SESSION] session={session_id}, active_task={active_task is not None}, "
+        f"msg_count={len(messages)}, last_role={messages[-1].get('role') if messages else 'none'}"
+    )
     if active_task:
         # Filter out the in-progress assistant message from the session response.
         # The client will receive the complete assistant response through the SSE
@@ -301,10 +316,9 @@ async def stream_chat_post(
         f"user={user_id}, message_len={len(request.message)}",
         extra={"json_fields": log_meta},
     )
-
     session = await _validate_and_get_session(session_id, user_id)
     logger.info(
-        f"[TIMING] session validated in {(time.perf_counter() - stream_start_time)*1000:.1f}ms",
+        f"[TIMING] session validated in {(time.perf_counter() - stream_start_time) * 1000:.1f}ms",
         extra={
             "json_fields": {
                 **log_meta,
@@ -313,6 +327,25 @@ async def stream_chat_post(
         },
     )
 
+    # Atomically append user message to session BEFORE creating task to avoid
+    # race condition where GET_SESSION sees task as "running" but message isn't
+    # saved yet.  append_and_save_message re-fetches inside a lock to prevent
+    # message loss from concurrent requests.
+    if request.message:
+        message = ChatMessage(
+            role="user" if request.is_user_message else "assistant",
+            content=request.message,
+        )
+        if request.is_user_message:
+            track_user_message(
+                user_id=user_id,
+                session_id=session_id,
+                message_length=len(request.message),
+            )
+        logger.info(f"[STREAM] Saving user message to session {session_id}")
+        session = await append_and_save_message(session_id, message)
+        logger.info(f"[STREAM] User message saved for session {session_id}")
+
     # Create a task in the stream registry for reconnection support
     task_id = str(uuid_module.uuid4())
     operation_id = str(uuid_module.uuid4())
@@ -328,7 +361,7 @@ async def stream_chat_post(
         operation_id=operation_id,
     )
     logger.info(
-        f"[TIMING] create_task completed in {(time.perf_counter() - task_create_start)*1000:.1f}ms",
+        f"[TIMING] create_task completed in {(time.perf_counter() - task_create_start) * 1000:.1f}ms",
         extra={
             "json_fields": {
                 **log_meta,
@@ -349,15 +382,47 @@ async def stream_chat_post(
         first_chunk_time, ttfc = None, None
         chunk_count = 0
         try:
-            async for chunk in chat_service.stream_chat_completion(
+            # Emit a start event with task_id for reconnection
+            start_chunk = StreamStart(messageId=task_id, taskId=task_id)
+            await stream_registry.publish_chunk(task_id, start_chunk)
+            logger.info(
+                f"[TIMING] StreamStart published at {(time_module.perf_counter() - gen_start_time) * 1000:.1f}ms",
+                extra={
+                    "json_fields": {
+                        **log_meta,
+                        "elapsed_ms": (time_module.perf_counter() - gen_start_time)
+                        * 1000,
+                    }
+                },
+            )
+
+            # Choose service based on LaunchDarkly flag (falls back to config default)
+            use_sdk = await is_feature_enabled(
+                Flag.COPILOT_SDK,
+                user_id or "anonymous",
+                default=config.use_claude_agent_sdk,
+            )
+            stream_fn = (
+                sdk_service.stream_chat_completion_sdk
+                if use_sdk
+                else chat_service.stream_chat_completion
+            )
+            logger.info(
+                f"[TIMING] Calling {'sdk' if use_sdk else 'standard'} stream_chat_completion",
+                extra={"json_fields": log_meta},
+            )
+            # Pass message=None since we already added it to the session above
+            async for chunk in stream_fn(
                 session_id,
-                request.message,
+                None,  # Message already in session
                 is_user_message=request.is_user_message,
                 user_id=user_id,
-                session=session,  # Pass pre-fetched session to avoid double-fetch
+                session=session,  # Pass session with message already added
                 context=request.context,
-                _task_id=task_id,  # Pass task_id so service emits start with taskId for reconnection
             ):
+                # Skip duplicate StreamStart — we already published one above
+                if isinstance(chunk, StreamStart):
+                    continue
                 chunk_count += 1
                 if first_chunk_time is None:
                     first_chunk_time = time_module.perf_counter()
@@ -378,7 +443,7 @@ async def stream_chat_post(
             gen_end_time = time_module.perf_counter()
             total_time = (gen_end_time - gen_start_time) * 1000
             logger.info(
-                f"[TIMING] run_ai_generation FINISHED in {total_time/1000:.1f}s; "
+                f"[TIMING] run_ai_generation FINISHED in {total_time / 1000:.1f}s; "
                 f"task={task_id}, session={session_id}, "
                 f"ttfc={ttfc or -1:.2f}s, n_chunks={chunk_count}",
                 extra={
@@ -405,6 +470,17 @@ async def stream_chat_post(
                     }
                 },
             )
+            # Publish a StreamError so the frontend can display an error message
+            try:
+                await stream_registry.publish_chunk(
+                    task_id,
+                    StreamError(
+                        errorText="An error occurred. Please try again.",
+                        code="stream_error",
+                    ),
+                )
+            except Exception:
+                pass  # Best-effort; mark_task_completed will publish StreamFinish
             await stream_registry.mark_task_completed(task_id, "failed")
 
     # Start the AI generation in a background task
@@ -507,8 +583,14 @@ async def stream_chat_post(
                     "json_fields": {**log_meta, "elapsed_ms": elapsed, "error": str(e)}
                 },
             )
+            # Surface error to frontend so it doesn't appear stuck
+            yield StreamError(
+                errorText="An error occurred. Please try again.",
+                code="stream_error",
+            ).to_sse()
+            yield StreamFinish().to_sse()
         finally:
-            # Unsubscribe when client disconnects or stream ends to prevent resource leak
+            # Unsubscribe when client disconnects or stream ends
             if subscriber_queue is not None:
                 try:
                     await stream_registry.unsubscribe_from_task(
@@ -752,8 +834,6 @@ async def stream_task(
         )
 
     async def event_generator() -> AsyncGenerator[str, None]:
-        import asyncio
-
         heartbeat_interval = 15.0  # Send heartbeat every 15 seconds
         try:
             while True:

autogpt_platform/backend/backend/api/features/chat/sdk/__init__.py

@@ -0,0 +1,14 @@
+"""Claude Agent SDK integration for CoPilot.
+
+This module provides the integration layer between the Claude Agent SDK
+and the existing CoPilot tool system, enabling drop-in replacement of
+the current LLM orchestration with the battle-tested Claude Agent SDK.
+"""
+
+from .service import stream_chat_completion_sdk
+from .tool_adapter import create_copilot_mcp_server
+
+__all__ = [
+    "stream_chat_completion_sdk",
+    "create_copilot_mcp_server",
+]

autogpt_platform/backend/backend/api/features/chat/sdk/response_adapter.py

@@ -0,0 +1,203 @@
+"""Response adapter for converting Claude Agent SDK messages to Vercel AI SDK format.
+
+This module provides the adapter layer that converts streaming messages from
+the Claude Agent SDK into the Vercel AI SDK UI Stream Protocol format that
+the frontend expects.
+"""
+
+import json
+import logging
+import uuid
+
+from claude_agent_sdk import (
+    AssistantMessage,
+    Message,
+    ResultMessage,
+    SystemMessage,
+    TextBlock,
+    ToolResultBlock,
+    ToolUseBlock,
+    UserMessage,
+)
+
+from backend.api.features.chat.response_model import (
+    StreamBaseResponse,
+    StreamError,
+    StreamFinish,
+    StreamFinishStep,
+    StreamStart,
+    StreamStartStep,
+    StreamTextDelta,
+    StreamTextEnd,
+    StreamTextStart,
+    StreamToolInputAvailable,
+    StreamToolInputStart,
+    StreamToolOutputAvailable,
+)
+from backend.api.features.chat.sdk.tool_adapter import (
+    MCP_TOOL_PREFIX,
+    pop_pending_tool_output,
+)
+
+logger = logging.getLogger(__name__)
+
+
+class SDKResponseAdapter:
+    """Adapter for converting Claude Agent SDK messages to Vercel AI SDK format.
+
+    This class maintains state during a streaming session to properly track
+    text blocks, tool calls, and message lifecycle.
+    """
+
+    def __init__(self, message_id: str | None = None):
+        self.message_id = message_id or str(uuid.uuid4())
+        self.text_block_id = str(uuid.uuid4())
+        self.has_started_text = False
+        self.has_ended_text = False
+        self.current_tool_calls: dict[str, dict[str, str]] = {}
+        self.task_id: str | None = None
+        self.step_open = False
+
+    def set_task_id(self, task_id: str) -> None:
+        """Set the task ID for reconnection support."""
+        self.task_id = task_id
+
+    def convert_message(self, sdk_message: Message) -> list[StreamBaseResponse]:
+        """Convert a single SDK message to Vercel AI SDK format."""
+        responses: list[StreamBaseResponse] = []
+
+        if isinstance(sdk_message, SystemMessage):
+            if sdk_message.subtype == "init":
+                responses.append(
+                    StreamStart(messageId=self.message_id, taskId=self.task_id)
+                )
+                # Open the first step (matches non-SDK: StreamStart then StreamStartStep)
+                responses.append(StreamStartStep())
+                self.step_open = True
+
+        elif isinstance(sdk_message, AssistantMessage):
+            # After tool results, the SDK sends a new AssistantMessage for the
+            # next LLM turn. Open a new step if the previous one was closed.
+            if not self.step_open:
+                responses.append(StreamStartStep())
+                self.step_open = True
+
+            for block in sdk_message.content:
+                if isinstance(block, TextBlock):
+                    if block.text:
+                        self._ensure_text_started(responses)
+                        responses.append(
+                            StreamTextDelta(id=self.text_block_id, delta=block.text)
+                        )
+
+                elif isinstance(block, ToolUseBlock):
+                    self._end_text_if_open(responses)
+
+                    # Strip MCP prefix so frontend sees "find_block"
+                    # instead of "mcp__copilot__find_block".
+                    tool_name = block.name.removeprefix(MCP_TOOL_PREFIX)
+
+                    responses.append(
+                        StreamToolInputStart(toolCallId=block.id, toolName=tool_name)
+                    )
+                    responses.append(
+                        StreamToolInputAvailable(
+                            toolCallId=block.id,
+                            toolName=tool_name,
+                            input=block.input,
+                        )
+                    )
+                    self.current_tool_calls[block.id] = {"name": tool_name}
+
+        elif isinstance(sdk_message, UserMessage):
+            # UserMessage carries tool results back from tool execution.
+            content = sdk_message.content
+            blocks = content if isinstance(content, list) else []
+            for block in blocks:
+                if isinstance(block, ToolResultBlock) and block.tool_use_id:
+                    tool_info = self.current_tool_calls.get(block.tool_use_id, {})
+                    tool_name = tool_info.get("name", "unknown")
+
+                    # Prefer the stashed full output over the SDK's
+                    # (potentially truncated) ToolResultBlock content.
+                    # The SDK truncates large results, writing them to disk,
+                    # which breaks frontend widget parsing.
+                    output = pop_pending_tool_output(tool_name) or (
+                        _extract_tool_output(block.content)
+                    )
+
+                    responses.append(
+                        StreamToolOutputAvailable(
+                            toolCallId=block.tool_use_id,
+                            toolName=tool_name,
+                            output=output,
+                            success=not (block.is_error or False),
+                        )
+                    )
+
+            # Close the current step after tool results — the next
+            # AssistantMessage will open a new step for the continuation.
+            if self.step_open:
+                responses.append(StreamFinishStep())
+                self.step_open = False
+
+        elif isinstance(sdk_message, ResultMessage):
+            self._end_text_if_open(responses)
+            # Close the step before finishing.
+            if self.step_open:
+                responses.append(StreamFinishStep())
+                self.step_open = False
+
+            if sdk_message.subtype == "success":
+                responses.append(StreamFinish())
+            elif sdk_message.subtype in ("error", "error_during_execution"):
+                error_msg = getattr(sdk_message, "result", None) or "Unknown error"
+                responses.append(
+                    StreamError(errorText=str(error_msg), code="sdk_error")
+                )
+                responses.append(StreamFinish())
+            else:
+                logger.warning(
+                    f"Unexpected ResultMessage subtype: {sdk_message.subtype}"
+                )
+                responses.append(StreamFinish())
+
+        else:
+            logger.debug(f"Unhandled SDK message type: {type(sdk_message).__name__}")
+
+        return responses
+
+    def _ensure_text_started(self, responses: list[StreamBaseResponse]) -> None:
+        """Start (or restart) a text block if needed."""
+        if not self.has_started_text or self.has_ended_text:
+            if self.has_ended_text:
+                self.text_block_id = str(uuid.uuid4())
+                self.has_ended_text = False
+            responses.append(StreamTextStart(id=self.text_block_id))
+            self.has_started_text = True
+
+    def _end_text_if_open(self, responses: list[StreamBaseResponse]) -> None:
+        """End the current text block if one is open."""
+        if self.has_started_text and not self.has_ended_text:
+            responses.append(StreamTextEnd(id=self.text_block_id))
+            self.has_ended_text = True
+
+
+def _extract_tool_output(content: str | list[dict[str, str]] | None) -> str:
+    """Extract a string output from a ToolResultBlock's content field."""
+    if isinstance(content, str):
+        return content
+    if isinstance(content, list):
+        parts = [item.get("text", "") for item in content if item.get("type") == "text"]
+        if parts:
+            return "".join(parts)
+        try:
+            return json.dumps(content)
+        except (TypeError, ValueError):
+            return str(content)
+    if content is None:
+        return ""
+    try:
+        return json.dumps(content)
+    except (TypeError, ValueError):
+        return str(content)

autogpt_platform/backend/backend/api/features/chat/sdk/response_adapter_test.py

@@ -0,0 +1,366 @@
+"""Unit tests for the SDK response adapter."""
+
+from claude_agent_sdk import (
+    AssistantMessage,
+    ResultMessage,
+    SystemMessage,
+    TextBlock,
+    ToolResultBlock,
+    ToolUseBlock,
+    UserMessage,
+)
+
+from backend.api.features.chat.response_model import (
+    StreamBaseResponse,
+    StreamError,
+    StreamFinish,
+    StreamFinishStep,
+    StreamStart,
+    StreamStartStep,
+    StreamTextDelta,
+    StreamTextEnd,
+    StreamTextStart,
+    StreamToolInputAvailable,
+    StreamToolInputStart,
+    StreamToolOutputAvailable,
+)
+
+from .response_adapter import SDKResponseAdapter
+from .tool_adapter import MCP_TOOL_PREFIX
+
+
+def _adapter() -> SDKResponseAdapter:
+    a = SDKResponseAdapter(message_id="msg-1")
+    a.set_task_id("task-1")
+    return a
+
+
+# -- SystemMessage -----------------------------------------------------------
+
+
+def test_system_init_emits_start_and_step():
+    adapter = _adapter()
+    results = adapter.convert_message(SystemMessage(subtype="init", data={}))
+    assert len(results) == 2
+    assert isinstance(results[0], StreamStart)
+    assert results[0].messageId == "msg-1"
+    assert results[0].taskId == "task-1"
+    assert isinstance(results[1], StreamStartStep)
+
+
+def test_system_non_init_emits_nothing():
+    adapter = _adapter()
+    results = adapter.convert_message(SystemMessage(subtype="other", data={}))
+    assert results == []
+
+
+# -- AssistantMessage with TextBlock -----------------------------------------
+
+
+def test_text_block_emits_step_start_and_delta():
+    adapter = _adapter()
+    msg = AssistantMessage(content=[TextBlock(text="hello")], model="test")
+    results = adapter.convert_message(msg)
+    assert len(results) == 3
+    assert isinstance(results[0], StreamStartStep)
+    assert isinstance(results[1], StreamTextStart)
+    assert isinstance(results[2], StreamTextDelta)
+    assert results[2].delta == "hello"
+
+
+def test_empty_text_block_emits_only_step():
+    adapter = _adapter()
+    msg = AssistantMessage(content=[TextBlock(text="")], model="test")
+    results = adapter.convert_message(msg)
+    # Empty text skipped, but step still opens
+    assert len(results) == 1
+    assert isinstance(results[0], StreamStartStep)
+
+
+def test_multiple_text_deltas_reuse_block_id():
+    adapter = _adapter()
+    msg1 = AssistantMessage(content=[TextBlock(text="a")], model="test")
+    msg2 = AssistantMessage(content=[TextBlock(text="b")], model="test")
+    r1 = adapter.convert_message(msg1)
+    r2 = adapter.convert_message(msg2)
+    # First gets step+start+delta, second only delta (block & step already started)
+    assert len(r1) == 3
+    assert isinstance(r1[0], StreamStartStep)
+    assert isinstance(r1[1], StreamTextStart)
+    assert len(r2) == 1
+    assert isinstance(r2[0], StreamTextDelta)
+    assert r1[1].id == r2[0].id  # same block ID
+
+
+# -- AssistantMessage with ToolUseBlock --------------------------------------
+
+
+def test_tool_use_emits_input_start_and_available():
+    """Tool names arrive with MCP prefix and should be stripped for the frontend."""
+    adapter = _adapter()
+    msg = AssistantMessage(
+        content=[
+            ToolUseBlock(
+                id="tool-1",
+                name=f"{MCP_TOOL_PREFIX}find_agent",
+                input={"q": "x"},
+            )
+        ],
+        model="test",
+    )
+    results = adapter.convert_message(msg)
+    assert len(results) == 3
+    assert isinstance(results[0], StreamStartStep)
+    assert isinstance(results[1], StreamToolInputStart)
+    assert results[1].toolCallId == "tool-1"
+    assert results[1].toolName == "find_agent"  # prefix stripped
+    assert isinstance(results[2], StreamToolInputAvailable)
+    assert results[2].toolName == "find_agent"  # prefix stripped
+    assert results[2].input == {"q": "x"}
+
+
+def test_text_then_tool_ends_text_block():
+    adapter = _adapter()
+    text_msg = AssistantMessage(content=[TextBlock(text="thinking...")], model="test")
+    tool_msg = AssistantMessage(
+        content=[ToolUseBlock(id="t1", name=f"{MCP_TOOL_PREFIX}tool", input={})],
+        model="test",
+    )
+    adapter.convert_message(text_msg)  # opens step + text
+    results = adapter.convert_message(tool_msg)
+    # Step already open, so: TextEnd, ToolInputStart, ToolInputAvailable
+    assert len(results) == 3
+    assert isinstance(results[0], StreamTextEnd)
+    assert isinstance(results[1], StreamToolInputStart)
+
+
+# -- UserMessage with ToolResultBlock ----------------------------------------
+
+
+def test_tool_result_emits_output_and_finish_step():
+    adapter = _adapter()
+    # First register the tool call (opens step) — SDK sends prefixed name
+    tool_msg = AssistantMessage(
+        content=[ToolUseBlock(id="t1", name=f"{MCP_TOOL_PREFIX}find_agent", input={})],
+        model="test",
+    )
+    adapter.convert_message(tool_msg)
+
+    # Now send tool result
+    result_msg = UserMessage(
+        content=[ToolResultBlock(tool_use_id="t1", content="found 3 agents")]
+    )
+    results = adapter.convert_message(result_msg)
+    assert len(results) == 2
+    assert isinstance(results[0], StreamToolOutputAvailable)
+    assert results[0].toolCallId == "t1"
+    assert results[0].toolName == "find_agent"  # prefix stripped
+    assert results[0].output == "found 3 agents"
+    assert results[0].success is True
+    assert isinstance(results[1], StreamFinishStep)
+
+
+def test_tool_result_error():
+    adapter = _adapter()
+    adapter.convert_message(
+        AssistantMessage(
+            content=[
+                ToolUseBlock(id="t1", name=f"{MCP_TOOL_PREFIX}run_agent", input={})
+            ],
+            model="test",
+        )
+    )
+    result_msg = UserMessage(
+        content=[ToolResultBlock(tool_use_id="t1", content="timeout", is_error=True)]
+    )
+    results = adapter.convert_message(result_msg)
+    assert isinstance(results[0], StreamToolOutputAvailable)
+    assert results[0].success is False
+    assert isinstance(results[1], StreamFinishStep)
+
+
+def test_tool_result_list_content():
+    adapter = _adapter()
+    adapter.convert_message(
+        AssistantMessage(
+            content=[ToolUseBlock(id="t1", name=f"{MCP_TOOL_PREFIX}tool", input={})],
+            model="test",
+        )
+    )
+    result_msg = UserMessage(
+        content=[
+            ToolResultBlock(
+                tool_use_id="t1",
+                content=[
+                    {"type": "text", "text": "line1"},
+                    {"type": "text", "text": "line2"},
+                ],
+            )
+        ]
+    )
+    results = adapter.convert_message(result_msg)
+    assert isinstance(results[0], StreamToolOutputAvailable)
+    assert results[0].output == "line1line2"
+    assert isinstance(results[1], StreamFinishStep)
+
+
+def test_string_user_message_ignored():
+    """A plain string UserMessage (not tool results) produces no output."""
+    adapter = _adapter()
+    results = adapter.convert_message(UserMessage(content="hello"))
+    assert results == []
+
+
+# -- ResultMessage -----------------------------------------------------------
+
+
+def test_result_success_emits_finish_step_and_finish():
+    adapter = _adapter()
+    # Start some text first (opens step)
+    adapter.convert_message(
+        AssistantMessage(content=[TextBlock(text="done")], model="test")
+    )
+    msg = ResultMessage(
+        subtype="success",
+        duration_ms=100,
+        duration_api_ms=50,
+        is_error=False,
+        num_turns=1,
+        session_id="s1",
+    )
+    results = adapter.convert_message(msg)
+    # TextEnd + FinishStep + StreamFinish
+    assert len(results) == 3
+    assert isinstance(results[0], StreamTextEnd)
+    assert isinstance(results[1], StreamFinishStep)
+    assert isinstance(results[2], StreamFinish)
+
+
+def test_result_error_emits_error_and_finish():
+    adapter = _adapter()
+    msg = ResultMessage(
+        subtype="error",
+        duration_ms=100,
+        duration_api_ms=50,
+        is_error=True,
+        num_turns=0,
+        session_id="s1",
+        result="API rate limited",
+    )
+    results = adapter.convert_message(msg)
+    # No step was open, so no FinishStep — just Error + Finish
+    assert len(results) == 2
+    assert isinstance(results[0], StreamError)
+    assert "API rate limited" in results[0].errorText
+    assert isinstance(results[1], StreamFinish)
+
+
+# -- Text after tools (new block ID) ----------------------------------------
+
+
+def test_text_after_tool_gets_new_block_id():
+    adapter = _adapter()
+    # Text -> Tool -> ToolResult -> Text should get a new text block ID and step
+    adapter.convert_message(
+        AssistantMessage(content=[TextBlock(text="before")], model="test")
+    )
+    adapter.convert_message(
+        AssistantMessage(
+            content=[ToolUseBlock(id="t1", name=f"{MCP_TOOL_PREFIX}tool", input={})],
+            model="test",
+        )
+    )
+    # Send tool result (closes step)
+    adapter.convert_message(
+        UserMessage(content=[ToolResultBlock(tool_use_id="t1", content="ok")])
+    )
+    results = adapter.convert_message(
+        AssistantMessage(content=[TextBlock(text="after")], model="test")
+    )
+    # Should get StreamStartStep (new step) + StreamTextStart (new block) + StreamTextDelta
+    assert len(results) == 3
+    assert isinstance(results[0], StreamStartStep)
+    assert isinstance(results[1], StreamTextStart)
+    assert isinstance(results[2], StreamTextDelta)
+    assert results[2].delta == "after"
+
+
+# -- Full conversation flow --------------------------------------------------
+
+
+def test_full_conversation_flow():
+    """Simulate a complete conversation: init -> text -> tool -> result -> text -> finish."""
+    adapter = _adapter()
+    all_responses: list[StreamBaseResponse] = []
+
+    # 1. Init
+    all_responses.extend(
+        adapter.convert_message(SystemMessage(subtype="init", data={}))
+    )
+    # 2. Assistant text
+    all_responses.extend(
+        adapter.convert_message(
+            AssistantMessage(content=[TextBlock(text="Let me search")], model="test")
+        )
+    )
+    # 3. Tool use
+    all_responses.extend(
+        adapter.convert_message(
+            AssistantMessage(
+                content=[
+                    ToolUseBlock(
+                        id="t1",
+                        name=f"{MCP_TOOL_PREFIX}find_agent",
+                        input={"query": "email"},
+                    )
+                ],
+                model="test",
+            )
+        )
+    )
+    # 4. Tool result
+    all_responses.extend(
+        adapter.convert_message(
+            UserMessage(
+                content=[ToolResultBlock(tool_use_id="t1", content="Found 2 agents")]
+            )
+        )
+    )
+    # 5. More text
+    all_responses.extend(
+        adapter.convert_message(
+            AssistantMessage(content=[TextBlock(text="I found 2")], model="test")
+        )
+    )
+    # 6. Result
+    all_responses.extend(
+        adapter.convert_message(
+            ResultMessage(
+                subtype="success",
+                duration_ms=500,
+                duration_api_ms=400,
+                is_error=False,
+                num_turns=2,
+                session_id="s1",
+            )
+        )
+    )
+
+    types = [type(r).__name__ for r in all_responses]
+    assert types == [
+        "StreamStart",
+        "StreamStartStep",  # step 1: text + tool call
+        "StreamTextStart",
+        "StreamTextDelta",  # "Let me search"
+        "StreamTextEnd",  # closed before tool
+        "StreamToolInputStart",
+        "StreamToolInputAvailable",
+        "StreamToolOutputAvailable",  # tool result
+        "StreamFinishStep",  # step 1 closed after tool result
+        "StreamStartStep",  # step 2: continuation text
+        "StreamTextStart",  # new block after tool
+        "StreamTextDelta",  # "I found 2"
+        "StreamTextEnd",  # closed by result
+        "StreamFinishStep",  # step 2 closed
+        "StreamFinish",
+    ]

autogpt_platform/backend/backend/api/features/chat/sdk/security_hooks.py

@@ -0,0 +1,335 @@
+"""Security hooks for Claude Agent SDK integration.
+
+This module provides security hooks that validate tool calls before execution,
+ensuring multi-user isolation and preventing unauthorized operations.
+"""
+
+import json
+import logging
+import os
+import re
+from collections.abc import Callable
+from typing import Any, cast
+
+from backend.api.features.chat.sdk.tool_adapter import MCP_TOOL_PREFIX
+
+logger = logging.getLogger(__name__)
+
+# Tools that are blocked entirely (CLI/system access).
+# "Bash" (capital) is the SDK built-in — it's NOT in allowed_tools but blocked
+# here as defence-in-depth.  The agent uses mcp__copilot__bash_exec instead,
+# which has kernel-level network isolation (unshare --net).
+BLOCKED_TOOLS = {
+    "Bash",
+    "bash",
+    "shell",
+    "exec",
+    "terminal",
+    "command",
+}
+
+# Tools allowed only when their path argument stays within the SDK workspace.
+# The SDK uses these to handle oversized tool results (writes to tool-results/
+# files, then reads them back) and for workspace file operations.
+WORKSPACE_SCOPED_TOOLS = {"Read", "Write", "Edit", "Glob", "Grep"}
+
+# Dangerous patterns in tool inputs
+DANGEROUS_PATTERNS = [
+    r"sudo",
+    r"rm\s+-rf",
+    r"dd\s+if=",
+    r"/etc/passwd",
+    r"/etc/shadow",
+    r"chmod\s+777",
+    r"curl\s+.*\|.*sh",
+    r"wget\s+.*\|.*sh",
+    r"eval\s*\(",
+    r"exec\s*\(",
+    r"__import__",
+    r"os\.system",
+    r"subprocess",
+]
+
+
+def _deny(reason: str) -> dict[str, Any]:
+    """Return a hook denial response."""
+    return {
+        "hookSpecificOutput": {
+            "hookEventName": "PreToolUse",
+            "permissionDecision": "deny",
+            "permissionDecisionReason": reason,
+        }
+    }
+
+
+def _validate_workspace_path(
+    tool_name: str, tool_input: dict[str, Any], sdk_cwd: str | None
+) -> dict[str, Any]:
+    """Validate that a workspace-scoped tool only accesses allowed paths.
+
+    Allowed directories:
+    - The SDK working directory (``/tmp/copilot-<session>/``)
+    - The SDK tool-results directory (``~/.claude/projects/…/tool-results/``)
+    """
+    path = tool_input.get("file_path") or tool_input.get("path") or ""
+    if not path:
+        # Glob/Grep without a path default to cwd which is already sandboxed
+        return {}
+
+    # Resolve relative paths against sdk_cwd (the SDK sets cwd so the LLM
+    # naturally uses relative paths like "test.txt" instead of absolute ones).
+    # Tilde paths (~/) are home-dir references, not relative — expand first.
+    if path.startswith("~"):
+        resolved = os.path.realpath(os.path.expanduser(path))
+    elif not os.path.isabs(path) and sdk_cwd:
+        resolved = os.path.realpath(os.path.join(sdk_cwd, path))
+    else:
+        resolved = os.path.realpath(path)
+
+    # Allow access within the SDK working directory
+    if sdk_cwd:
+        norm_cwd = os.path.realpath(sdk_cwd)
+        if resolved.startswith(norm_cwd + os.sep) or resolved == norm_cwd:
+            return {}
+
+    # Allow access to ~/.claude/projects/*/tool-results/ (big tool results)
+    claude_dir = os.path.realpath(os.path.expanduser("~/.claude/projects"))
+    tool_results_seg = os.sep + "tool-results" + os.sep
+    if resolved.startswith(claude_dir + os.sep) and tool_results_seg in resolved:
+        return {}
+
+    logger.warning(
+        f"Blocked {tool_name} outside workspace: {path} (resolved={resolved})"
+    )
+    workspace_hint = f" Allowed workspace: {sdk_cwd}" if sdk_cwd else ""
+    return _deny(
+        f"[SECURITY] Tool '{tool_name}' can only access files within the workspace "
+        f"directory.{workspace_hint} "
+        "This is enforced by the platform and cannot be bypassed."
+    )
+
+
+def _validate_tool_access(
+    tool_name: str, tool_input: dict[str, Any], sdk_cwd: str | None = None
+) -> dict[str, Any]:
+    """Validate that a tool call is allowed.
+
+    Returns:
+        Empty dict to allow, or dict with hookSpecificOutput to deny
+    """
+    # Block forbidden tools
+    if tool_name in BLOCKED_TOOLS:
+        logger.warning(f"Blocked tool access attempt: {tool_name}")
+        return _deny(
+            f"[SECURITY] Tool '{tool_name}' is blocked for security. "
+            "This is enforced by the platform and cannot be bypassed. "
+            "Use the CoPilot-specific MCP tools instead."
+        )
+
+    # Workspace-scoped tools: allowed only within the SDK workspace directory
+    if tool_name in WORKSPACE_SCOPED_TOOLS:
+        return _validate_workspace_path(tool_name, tool_input, sdk_cwd)
+
+    # Check for dangerous patterns in tool input
+    # Use json.dumps for predictable format (str() produces Python repr)
+    input_str = json.dumps(tool_input) if tool_input else ""
+
+    for pattern in DANGEROUS_PATTERNS:
+        if re.search(pattern, input_str, re.IGNORECASE):
+            logger.warning(
+                f"Blocked dangerous pattern in tool input: {pattern} in {tool_name}"
+            )
+            return _deny(
+                "[SECURITY] Input contains a blocked pattern. "
+                "This is enforced by the platform and cannot be bypassed."
+            )
+
+    return {}
+
+
+def _validate_user_isolation(
+    tool_name: str, tool_input: dict[str, Any], user_id: str | None
+) -> dict[str, Any]:
+    """Validate that tool calls respect user isolation."""
+    # For workspace file tools, ensure path doesn't escape
+    if "workspace" in tool_name.lower():
+        path = tool_input.get("path", "") or tool_input.get("file_path", "")
+        if path:
+            # Check for path traversal
+            if ".." in path or path.startswith("/"):
+                logger.warning(
+                    f"Blocked path traversal attempt: {path} by user {user_id}"
+                )
+                return {
+                    "hookSpecificOutput": {
+                        "hookEventName": "PreToolUse",
+                        "permissionDecision": "deny",
+                        "permissionDecisionReason": "Path traversal not allowed",
+                    }
+                }
+
+    return {}
+
+
+def create_security_hooks(
+    user_id: str | None,
+    sdk_cwd: str | None = None,
+    max_subtasks: int = 3,
+    on_stop: Callable[[str, str], None] | None = None,
+) -> dict[str, Any]:
+    """Create the security hooks configuration for Claude Agent SDK.
+
+    Includes security validation and observability hooks:
+    - PreToolUse: Security validation before tool execution
+    - PostToolUse: Log successful tool executions
+    - PostToolUseFailure: Log and handle failed tool executions
+    - PreCompact: Log context compaction events (SDK handles compaction automatically)
+    - Stop: Capture transcript path for stateless resume (when *on_stop* is provided)
+
+    Args:
+        user_id: Current user ID for isolation validation
+        sdk_cwd: SDK working directory for workspace-scoped tool validation
+        max_subtasks: Maximum Task (sub-agent) spawns allowed per session
+        on_stop: Callback ``(transcript_path, sdk_session_id)`` invoked when
+            the SDK finishes processing — used to read the JSONL transcript
+            before the CLI process exits.
+
+    Returns:
+        Hooks configuration dict for ClaudeAgentOptions
+    """
+    try:
+        from claude_agent_sdk import HookMatcher
+        from claude_agent_sdk.types import HookContext, HookInput, SyncHookJSONOutput
+
+        # Per-session counter for Task sub-agent spawns
+        task_spawn_count = 0
+
+        async def pre_tool_use_hook(
+            input_data: HookInput,
+            tool_use_id: str | None,
+            context: HookContext,
+        ) -> SyncHookJSONOutput:
+            """Combined pre-tool-use validation hook."""
+            nonlocal task_spawn_count
+            _ = context  # unused but required by signature
+            tool_name = cast(str, input_data.get("tool_name", ""))
+            tool_input = cast(dict[str, Any], input_data.get("tool_input", {}))
+
+            # Rate-limit Task (sub-agent) spawns per session
+            if tool_name == "Task":
+                task_spawn_count += 1
+                if task_spawn_count > max_subtasks:
+                    logger.warning(
+                        f"[SDK] Task limit reached ({max_subtasks}), user={user_id}"
+                    )
+                    return cast(
+                        SyncHookJSONOutput,
+                        _deny(
+                            f"Maximum {max_subtasks} sub-tasks per session. "
+                            "Please continue in the main conversation."
+                        ),
+                    )
+
+            # Strip MCP prefix for consistent validation
+            is_copilot_tool = tool_name.startswith(MCP_TOOL_PREFIX)
+            clean_name = tool_name.removeprefix(MCP_TOOL_PREFIX)
+
+            # Only block non-CoPilot tools; our MCP-registered tools
+            # (including Read for oversized results) are already sandboxed.
+            if not is_copilot_tool:
+                result = _validate_tool_access(clean_name, tool_input, sdk_cwd)
+                if result:
+                    return cast(SyncHookJSONOutput, result)
+
+            # Validate user isolation
+            result = _validate_user_isolation(clean_name, tool_input, user_id)
+            if result:
+                return cast(SyncHookJSONOutput, result)
+
+            logger.debug(f"[SDK] Tool start: {tool_name}, user={user_id}")
+            return cast(SyncHookJSONOutput, {})
+
+        async def post_tool_use_hook(
+            input_data: HookInput,
+            tool_use_id: str | None,
+            context: HookContext,
+        ) -> SyncHookJSONOutput:
+            """Log successful tool executions for observability."""
+            _ = context
+            tool_name = cast(str, input_data.get("tool_name", ""))
+            logger.debug(f"[SDK] Tool success: {tool_name}, tool_use_id={tool_use_id}")
+            return cast(SyncHookJSONOutput, {})
+
+        async def post_tool_failure_hook(
+            input_data: HookInput,
+            tool_use_id: str | None,
+            context: HookContext,
+        ) -> SyncHookJSONOutput:
+            """Log failed tool executions for debugging."""
+            _ = context
+            tool_name = cast(str, input_data.get("tool_name", ""))
+            error = input_data.get("error", "Unknown error")
+            logger.warning(
+                f"[SDK] Tool failed: {tool_name}, error={error}, "
+                f"user={user_id}, tool_use_id={tool_use_id}"
+            )
+            return cast(SyncHookJSONOutput, {})
+
+        async def pre_compact_hook(
+            input_data: HookInput,
+            tool_use_id: str | None,
+            context: HookContext,
+        ) -> SyncHookJSONOutput:
+            """Log when SDK triggers context compaction.
+
+            The SDK automatically compacts conversation history when it grows too large.
+            This hook provides visibility into when compaction happens.
+            """
+            _ = context, tool_use_id
+            trigger = input_data.get("trigger", "auto")
+            logger.info(
+                f"[SDK] Context compaction triggered: {trigger}, user={user_id}"
+            )
+            return cast(SyncHookJSONOutput, {})
+
+        # --- Stop hook: capture transcript path for stateless resume ---
+        async def stop_hook(
+            input_data: HookInput,
+            tool_use_id: str | None,
+            context: HookContext,
+        ) -> SyncHookJSONOutput:
+            """Capture transcript path when SDK finishes processing.
+
+            The Stop hook fires while the CLI process is still alive, giving us
+            a reliable window to read the JSONL transcript before SIGTERM.
+            """
+            _ = context, tool_use_id
+            transcript_path = cast(str, input_data.get("transcript_path", ""))
+            sdk_session_id = cast(str, input_data.get("session_id", ""))
+
+            if transcript_path and on_stop:
+                logger.info(
+                    f"[SDK] Stop hook: transcript_path={transcript_path}, "
+                    f"sdk_session_id={sdk_session_id[:12]}..."
+                )
+                on_stop(transcript_path, sdk_session_id)
+
+            return cast(SyncHookJSONOutput, {})
+
+        hooks: dict[str, Any] = {
+            "PreToolUse": [HookMatcher(matcher="*", hooks=[pre_tool_use_hook])],
+            "PostToolUse": [HookMatcher(matcher="*", hooks=[post_tool_use_hook])],
+            "PostToolUseFailure": [
+                HookMatcher(matcher="*", hooks=[post_tool_failure_hook])
+            ],
+            "PreCompact": [HookMatcher(matcher="*", hooks=[pre_compact_hook])],
+        }
+
+        if on_stop is not None:
+            hooks["Stop"] = [HookMatcher(matcher=None, hooks=[stop_hook])]
+
+        return hooks
+    except ImportError:
+        # Fallback for when SDK isn't available - return empty hooks
+        logger.warning("claude-agent-sdk not available, security hooks disabled")
+        return {}

autogpt_platform/backend/backend/api/features/chat/sdk/security_hooks_test.py

@@ -0,0 +1,165 @@
+"""Unit tests for SDK security hooks."""
+
+import os
+
+from .security_hooks import _validate_tool_access, _validate_user_isolation
+
+SDK_CWD = "/tmp/copilot-abc123"
+
+
+def _is_denied(result: dict) -> bool:
+    hook = result.get("hookSpecificOutput", {})
+    return hook.get("permissionDecision") == "deny"
+
+
+# -- Blocked tools -----------------------------------------------------------
+
+
+def test_blocked_tools_denied():
+    for tool in ("bash", "shell", "exec", "terminal", "command"):
+        result = _validate_tool_access(tool, {})
+        assert _is_denied(result), f"{tool} should be blocked"
+
+
+def test_unknown_tool_allowed():
+    result = _validate_tool_access("SomeCustomTool", {})
+    assert result == {}
+
+
+# -- Workspace-scoped tools --------------------------------------------------
+
+
+def test_read_within_workspace_allowed():
+    result = _validate_tool_access(
+        "Read", {"file_path": f"{SDK_CWD}/file.txt"}, sdk_cwd=SDK_CWD
+    )
+    assert result == {}
+
+
+def test_write_within_workspace_allowed():
+    result = _validate_tool_access(
+        "Write", {"file_path": f"{SDK_CWD}/output.json"}, sdk_cwd=SDK_CWD
+    )
+    assert result == {}
+
+
+def test_edit_within_workspace_allowed():
+    result = _validate_tool_access(
+        "Edit", {"file_path": f"{SDK_CWD}/src/main.py"}, sdk_cwd=SDK_CWD
+    )
+    assert result == {}
+
+
+def test_glob_within_workspace_allowed():
+    result = _validate_tool_access("Glob", {"path": f"{SDK_CWD}/src"}, sdk_cwd=SDK_CWD)
+    assert result == {}
+
+
+def test_grep_within_workspace_allowed():
+    result = _validate_tool_access("Grep", {"path": f"{SDK_CWD}/src"}, sdk_cwd=SDK_CWD)
+    assert result == {}
+
+
+def test_read_outside_workspace_denied():
+    result = _validate_tool_access(
+        "Read", {"file_path": "/etc/passwd"}, sdk_cwd=SDK_CWD
+    )
+    assert _is_denied(result)
+
+
+def test_write_outside_workspace_denied():
+    result = _validate_tool_access(
+        "Write", {"file_path": "/home/user/secrets.txt"}, sdk_cwd=SDK_CWD
+    )
+    assert _is_denied(result)
+
+
+def test_traversal_attack_denied():
+    result = _validate_tool_access(
+        "Read",
+        {"file_path": f"{SDK_CWD}/../../etc/passwd"},
+        sdk_cwd=SDK_CWD,
+    )
+    assert _is_denied(result)
+
+
+def test_no_path_allowed():
+    """Glob/Grep without a path argument defaults to cwd — should pass."""
+    result = _validate_tool_access("Glob", {}, sdk_cwd=SDK_CWD)
+    assert result == {}
+
+
+def test_read_no_cwd_denies_absolute():
+    """If no sdk_cwd is set, absolute paths are denied."""
+    result = _validate_tool_access("Read", {"file_path": "/tmp/anything"})
+    assert _is_denied(result)
+
+
+# -- Tool-results directory --------------------------------------------------
+
+
+def test_read_tool_results_allowed():
+    home = os.path.expanduser("~")
+    path = f"{home}/.claude/projects/-tmp-copilot-abc123/tool-results/12345.txt"
+    result = _validate_tool_access("Read", {"file_path": path}, sdk_cwd=SDK_CWD)
+    assert result == {}
+
+
+def test_read_claude_projects_without_tool_results_denied():
+    home = os.path.expanduser("~")
+    path = f"{home}/.claude/projects/-tmp-copilot-abc123/settings.json"
+    result = _validate_tool_access("Read", {"file_path": path}, sdk_cwd=SDK_CWD)
+    assert _is_denied(result)
+
+
+# -- Built-in Bash is blocked (use bash_exec MCP tool instead) ---------------
+
+
+def test_bash_builtin_always_blocked():
+    """SDK built-in Bash is blocked — bash_exec MCP tool with bubblewrap is used instead."""
+    result = _validate_tool_access("Bash", {"command": "echo hello"}, sdk_cwd=SDK_CWD)
+    assert _is_denied(result)
+
+
+# -- Dangerous patterns ------------------------------------------------------
+
+
+def test_dangerous_pattern_blocked():
+    result = _validate_tool_access("SomeTool", {"cmd": "sudo rm -rf /"})
+    assert _is_denied(result)
+
+
+def test_subprocess_pattern_blocked():
+    result = _validate_tool_access("SomeTool", {"code": "subprocess.run(...)"})
+    assert _is_denied(result)
+
+
+# -- User isolation ----------------------------------------------------------
+
+
+def test_workspace_path_traversal_blocked():
+    result = _validate_user_isolation(
+        "workspace_read", {"path": "../../../etc/shadow"}, user_id="user-1"
+    )
+    assert _is_denied(result)
+
+
+def test_workspace_absolute_path_blocked():
+    result = _validate_user_isolation(
+        "workspace_read", {"path": "/etc/passwd"}, user_id="user-1"
+    )
+    assert _is_denied(result)
+
+
+def test_workspace_normal_path_allowed():
+    result = _validate_user_isolation(
+        "workspace_read", {"path": "src/main.py"}, user_id="user-1"
+    )
+    assert result == {}
+
+
+def test_non_workspace_tool_passes_isolation():
+    result = _validate_user_isolation(
+        "find_agent", {"query": "email"}, user_id="user-1"
+    )
+    assert result == {}

autogpt_platform/backend/backend/api/features/chat/sdk/service.py

@@ -0,0 +1,751 @@
+"""Claude Agent SDK service layer for CoPilot chat completions."""
+
+import asyncio
+import json
+import logging
+import os
+import uuid
+from collections.abc import AsyncGenerator
+from dataclasses import dataclass
+from typing import Any
+
+from backend.util.exceptions import NotFoundError
+
+from .. import stream_registry
+from ..config import ChatConfig
+from ..model import (
+    ChatMessage,
+    ChatSession,
+    get_chat_session,
+    update_session_title,
+    upsert_chat_session,
+)
+from ..response_model import (
+    StreamBaseResponse,
+    StreamError,
+    StreamFinish,
+    StreamStart,
+    StreamTextDelta,
+    StreamToolInputAvailable,
+    StreamToolOutputAvailable,
+)
+from ..service import (
+    _build_system_prompt,
+    _execute_long_running_tool_with_streaming,
+    _generate_session_title,
+)
+from ..tools.models import OperationPendingResponse, OperationStartedResponse
+from ..tools.sandbox import WORKSPACE_PREFIX, make_session_path
+from ..tracking import track_user_message
+from .response_adapter import SDKResponseAdapter
+from .security_hooks import create_security_hooks
+from .tool_adapter import (
+    COPILOT_TOOL_NAMES,
+    LongRunningCallback,
+    create_copilot_mcp_server,
+    set_execution_context,
+)
+from .transcript import (
+    download_transcript,
+    read_transcript_file,
+    upload_transcript,
+    validate_transcript,
+    write_transcript_to_tempfile,
+)
+
+logger = logging.getLogger(__name__)
+config = ChatConfig()
+
+# Set to hold background tasks to prevent garbage collection
+_background_tasks: set[asyncio.Task[Any]] = set()
+
+
+@dataclass
+class CapturedTranscript:
+    """Info captured by the SDK Stop hook for stateless --resume."""
+
+    path: str = ""
+    sdk_session_id: str = ""
+
+    @property
+    def available(self) -> bool:
+        return bool(self.path)
+
+
+_SDK_CWD_PREFIX = WORKSPACE_PREFIX
+
+# Appended to the system prompt to inform the agent about available tools.
+# The SDK built-in Bash is NOT available — use mcp__copilot__bash_exec instead,
+# which has kernel-level network isolation (unshare --net).
+_SDK_TOOL_SUPPLEMENT = """
+
+## Tool notes
+
+- The SDK built-in Bash tool is NOT available.  Use the `bash_exec` MCP tool
+  for shell commands — it runs in a network-isolated sandbox.
+- **Shared workspace**: The SDK Read/Write tools and `bash_exec` share the
+  same working directory. Files created by one are readable by the other.
+  These files are **ephemeral** — they exist only for the current session.
+- **Persistent storage**: Use `write_workspace_file` / `read_workspace_file`
+  for files that should persist across sessions (stored in cloud storage).
+- Long-running tools (create_agent, edit_agent, etc.) are handled
+  asynchronously.  You will receive an immediate response; the actual result
+  is delivered to the user via a background stream.
+"""
+
+
+def _build_long_running_callback(user_id: str | None) -> LongRunningCallback:
+    """Build a callback that delegates long-running tools to the non-SDK infrastructure.
+
+    Long-running tools (create_agent, edit_agent, etc.) are delegated to the
+    existing background infrastructure: stream_registry (Redis Streams),
+    database persistence, and SSE reconnection.  This means results survive
+    page refreshes / pod restarts, and the frontend shows the proper loading
+    widget with progress updates.
+
+    The returned callback matches the ``LongRunningCallback`` signature:
+    ``(tool_name, args, session) -> MCP response dict``.
+    """
+
+    async def _callback(
+        tool_name: str, args: dict[str, Any], session: ChatSession
+    ) -> dict[str, Any]:
+        operation_id = str(uuid.uuid4())
+        task_id = str(uuid.uuid4())
+        tool_call_id = f"sdk-{uuid.uuid4().hex[:12]}"
+        session_id = session.session_id
+
+        # --- Build user-friendly messages (matches non-SDK service) ---
+        if tool_name == "create_agent":
+            desc = args.get("description", "")
+            desc_preview = (desc[:100] + "...") if len(desc) > 100 else desc
+            pending_msg = (
+                f"Creating your agent: {desc_preview}"
+                if desc_preview
+                else "Creating agent... This may take a few minutes."
+            )
+            started_msg = (
+                "Agent creation started. You can close this tab - "
+                "check your library in a few minutes."
+            )
+        elif tool_name == "edit_agent":
+            changes = args.get("changes", "")
+            changes_preview = (changes[:100] + "...") if len(changes) > 100 else changes
+            pending_msg = (
+                f"Editing agent: {changes_preview}"
+                if changes_preview
+                else "Editing agent... This may take a few minutes."
+            )
+            started_msg = (
+                "Agent edit started. You can close this tab - "
+                "check your library in a few minutes."
+            )
+        else:
+            pending_msg = f"Running {tool_name}... This may take a few minutes."
+            started_msg = (
+                f"{tool_name} started. You can close this tab - "
+                "check back in a few minutes."
+            )
+
+        # --- Register task in Redis for SSE reconnection ---
+        await stream_registry.create_task(
+            task_id=task_id,
+            session_id=session_id,
+            user_id=user_id,
+            tool_call_id=tool_call_id,
+            tool_name=tool_name,
+            operation_id=operation_id,
+        )
+
+        # --- Save OperationPendingResponse to chat history ---
+        pending_message = ChatMessage(
+            role="tool",
+            content=OperationPendingResponse(
+                message=pending_msg,
+                operation_id=operation_id,
+                tool_name=tool_name,
+            ).model_dump_json(),
+            tool_call_id=tool_call_id,
+        )
+        session.messages.append(pending_message)
+        await upsert_chat_session(session)
+
+        # --- Spawn background task (reuses non-SDK infrastructure) ---
+        bg_task = asyncio.create_task(
+            _execute_long_running_tool_with_streaming(
+                tool_name=tool_name,
+                parameters=args,
+                tool_call_id=tool_call_id,
+                operation_id=operation_id,
+                task_id=task_id,
+                session_id=session_id,
+                user_id=user_id,
+            )
+        )
+        _background_tasks.add(bg_task)
+        bg_task.add_done_callback(_background_tasks.discard)
+        await stream_registry.set_task_asyncio_task(task_id, bg_task)
+
+        logger.info(
+            f"[SDK] Long-running tool {tool_name} delegated to background "
+            f"(operation_id={operation_id}, task_id={task_id})"
+        )
+
+        # --- Return OperationStartedResponse as MCP tool result ---
+        # This flows through SDK → response adapter → frontend, triggering
+        # the loading widget with SSE reconnection support.
+        started_json = OperationStartedResponse(
+            message=started_msg,
+            operation_id=operation_id,
+            tool_name=tool_name,
+            task_id=task_id,
+        ).model_dump_json()
+
+        return {
+            "content": [{"type": "text", "text": started_json}],
+            "isError": False,
+        }
+
+    return _callback
+
+
+def _resolve_sdk_model() -> str | None:
+    """Resolve the model name for the Claude Agent SDK CLI.
+
+    Uses ``config.claude_agent_model`` if set, otherwise derives from
+    ``config.model`` by stripping the OpenRouter provider prefix (e.g.,
+    ``"anthropic/claude-opus-4.6"`` → ``"claude-opus-4.6"``).
+    """
+    if config.claude_agent_model:
+        return config.claude_agent_model
+    model = config.model
+    if "/" in model:
+        return model.split("/", 1)[1]
+    return model
+
+
+def _build_sdk_env() -> dict[str, str]:
+    """Build env vars for the SDK CLI process.
+
+    Routes API calls through OpenRouter (or a custom base_url) using
+    the same ``config.api_key`` / ``config.base_url`` as the non-SDK path.
+    This gives per-call token and cost tracking on the OpenRouter dashboard.
+
+    Only overrides ``ANTHROPIC_API_KEY`` when a valid proxy URL and auth
+    token are both present — otherwise returns an empty dict so the SDK
+    falls back to its default credentials.
+    """
+    env: dict[str, str] = {}
+    if config.api_key and config.base_url:
+        # Strip /v1 suffix — SDK expects the base URL without a version path
+        base = config.base_url.rstrip("/")
+        if base.endswith("/v1"):
+            base = base[:-3]
+        if not base or not base.startswith("http"):
+            # Invalid base_url — don't override SDK defaults
+            return env
+        env["ANTHROPIC_BASE_URL"] = base
+        env["ANTHROPIC_AUTH_TOKEN"] = config.api_key
+        # Must be explicitly empty so the CLI uses AUTH_TOKEN instead
+        env["ANTHROPIC_API_KEY"] = ""
+    return env
+
+
+def _make_sdk_cwd(session_id: str) -> str:
+    """Create a safe, session-specific working directory path.
+
+    Delegates to :func:`~backend.api.features.chat.tools.sandbox.make_session_path`
+    (single source of truth for path sanitization) and adds a defence-in-depth
+    assertion.
+    """
+    cwd = make_session_path(session_id)
+    # Defence-in-depth: normpath + startswith is a CodeQL-recognised sanitizer
+    cwd = os.path.normpath(cwd)
+    if not cwd.startswith(_SDK_CWD_PREFIX):
+        raise ValueError(f"SDK cwd escaped prefix: {cwd}")
+    return cwd
+
+
+def _cleanup_sdk_tool_results(cwd: str) -> None:
+    """Remove SDK tool-result files for a specific session working directory.
+
+    The SDK creates tool-result files under ~/.claude/projects/<encoded-cwd>/tool-results/.
+    We clean only the specific cwd's results to avoid race conditions between
+    concurrent sessions.
+
+    Security: cwd MUST be created by _make_sdk_cwd() which sanitizes session_id.
+    """
+    import shutil
+
+    # Validate cwd is under the expected prefix
+    normalized = os.path.normpath(cwd)
+    if not normalized.startswith(_SDK_CWD_PREFIX):
+        logger.warning(f"[SDK] Rejecting cleanup for path outside workspace: {cwd}")
+        return
+
+    # SDK encodes the cwd path by replacing '/' with '-'
+    encoded_cwd = normalized.replace("/", "-")
+
+    # Construct the project directory path (known-safe home expansion)
+    claude_projects = os.path.expanduser("~/.claude/projects")
+    project_dir = os.path.join(claude_projects, encoded_cwd)
+
+    # Security check 3: Validate project_dir is under ~/.claude/projects
+    project_dir = os.path.normpath(project_dir)
+    if not project_dir.startswith(claude_projects):
+        logger.warning(
+            f"[SDK] Rejecting cleanup for escaped project path: {project_dir}"
+        )
+        return
+
+    results_dir = os.path.join(project_dir, "tool-results")
+    if os.path.isdir(results_dir):
+        for filename in os.listdir(results_dir):
+            file_path = os.path.join(results_dir, filename)
+            try:
+                if os.path.isfile(file_path):
+                    os.remove(file_path)
+            except OSError:
+                pass
+
+    # Also clean up the temp cwd directory itself
+    try:
+        shutil.rmtree(normalized, ignore_errors=True)
+    except OSError:
+        pass
+
+
+async def _compress_conversation_history(
+    session: ChatSession,
+) -> list[ChatMessage]:
+    """Compress prior conversation messages if they exceed the token threshold.
+
+    Uses the shared compress_context() from prompt.py which supports:
+    - LLM summarization of old messages (keeps recent ones intact)
+    - Progressive content truncation as fallback
+    - Middle-out deletion as last resort
+
+    Returns the compressed prior messages (everything except the current message).
+    """
+    prior = session.messages[:-1]
+    if len(prior) < 2:
+        return prior
+
+    from backend.util.prompt import compress_context
+
+    # Convert ChatMessages to dicts for compress_context
+    messages_dict = []
+    for msg in prior:
+        msg_dict: dict[str, Any] = {"role": msg.role}
+        if msg.content:
+            msg_dict["content"] = msg.content
+        if msg.tool_calls:
+            msg_dict["tool_calls"] = msg.tool_calls
+        if msg.tool_call_id:
+            msg_dict["tool_call_id"] = msg.tool_call_id
+        messages_dict.append(msg_dict)
+
+    try:
+        import openai
+
+        async with openai.AsyncOpenAI(
+            api_key=config.api_key, base_url=config.base_url, timeout=30.0
+        ) as client:
+            result = await compress_context(
+                messages=messages_dict,
+                model=config.model,
+                client=client,
+            )
+    except Exception as e:
+        logger.warning(f"[SDK] Context compression with LLM failed: {e}")
+        # Fall back to truncation-only (no LLM summarization)
+        result = await compress_context(
+            messages=messages_dict,
+            model=config.model,
+            client=None,
+        )
+
+    if result.was_compacted:
+        logger.info(
+            f"[SDK] Context compacted: {result.original_token_count} -> "
+            f"{result.token_count} tokens "
+            f"({result.messages_summarized} summarized, "
+            f"{result.messages_dropped} dropped)"
+        )
+        # Convert compressed dicts back to ChatMessages
+        return [
+            ChatMessage(
+                role=m["role"],
+                content=m.get("content"),
+                tool_calls=m.get("tool_calls"),
+                tool_call_id=m.get("tool_call_id"),
+            )
+            for m in result.messages
+        ]
+
+    return prior
+
+
+def _format_conversation_context(messages: list[ChatMessage]) -> str | None:
+    """Format conversation messages into a context prefix for the user message.
+
+    Returns a string like:
+        <conversation_history>
+        User: hello
+        You responded: Hi! How can I help?
+        </conversation_history>
+
+    Returns None if there are no messages to format.
+    """
+    if not messages:
+        return None
+
+    lines: list[str] = []
+    for msg in messages:
+        if not msg.content:
+            continue
+        if msg.role == "user":
+            lines.append(f"User: {msg.content}")
+        elif msg.role == "assistant":
+            lines.append(f"You responded: {msg.content}")
+        # Skip tool messages — they're internal details
+
+    if not lines:
+        return None
+
+    return "<conversation_history>\n" + "\n".join(lines) + "\n</conversation_history>"
+
+
+async def stream_chat_completion_sdk(
+    session_id: str,
+    message: str | None = None,
+    tool_call_response: str | None = None,  # noqa: ARG001
+    is_user_message: bool = True,
+    user_id: str | None = None,
+    retry_count: int = 0,  # noqa: ARG001
+    session: ChatSession | None = None,
+    context: dict[str, str] | None = None,  # noqa: ARG001
+) -> AsyncGenerator[StreamBaseResponse, None]:
+    """Stream chat completion using Claude Agent SDK.
+
+    Drop-in replacement for stream_chat_completion with improved reliability.
+    """
+
+    if session is None:
+        session = await get_chat_session(session_id, user_id)
+
+    if not session:
+        raise NotFoundError(
+            f"Session {session_id} not found. Please create a new session first."
+        )
+
+    if message:
+        session.messages.append(
+            ChatMessage(
+                role="user" if is_user_message else "assistant", content=message
+            )
+        )
+        if is_user_message:
+            track_user_message(
+                user_id=user_id, session_id=session_id, message_length=len(message)
+            )
+
+    session = await upsert_chat_session(session)
+
+    # Generate title for new sessions (first user message)
+    if is_user_message and not session.title:
+        user_messages = [m for m in session.messages if m.role == "user"]
+        if len(user_messages) == 1:
+            first_message = user_messages[0].content or message or ""
+            if first_message:
+                task = asyncio.create_task(
+                    _update_title_async(session_id, first_message, user_id)
+                )
+                _background_tasks.add(task)
+                task.add_done_callback(_background_tasks.discard)
+
+    # Build system prompt (reuses non-SDK path with Langfuse support)
+    has_history = len(session.messages) > 1
+    system_prompt, _ = await _build_system_prompt(
+        user_id, has_conversation_history=has_history
+    )
+    system_prompt += _SDK_TOOL_SUPPLEMENT
+    message_id = str(uuid.uuid4())
+    task_id = str(uuid.uuid4())
+
+    yield StreamStart(messageId=message_id, taskId=task_id)
+
+    stream_completed = False
+    # Initialise sdk_cwd before the try so the finally can reference it
+    # even if _make_sdk_cwd raises (in that case it stays as "").
+    sdk_cwd = ""
+    use_resume = False
+
+    try:
+        # Use a session-specific temp dir to avoid cleanup race conditions
+        # between concurrent sessions.
+        sdk_cwd = _make_sdk_cwd(session_id)
+        os.makedirs(sdk_cwd, exist_ok=True)
+
+        set_execution_context(
+            user_id,
+            session,
+            long_running_callback=_build_long_running_callback(user_id),
+        )
+        try:
+            from claude_agent_sdk import ClaudeAgentOptions, ClaudeSDKClient
+
+            # Fail fast when no API credentials are available at all
+            sdk_env = _build_sdk_env()
+            if not sdk_env and not os.environ.get("ANTHROPIC_API_KEY"):
+                raise RuntimeError(
+                    "No API key configured. Set OPEN_ROUTER_API_KEY "
+                    "(or CHAT_API_KEY) for OpenRouter routing, "
+                    "or ANTHROPIC_API_KEY for direct Anthropic access."
+                )
+
+            mcp_server = create_copilot_mcp_server()
+
+            sdk_model = _resolve_sdk_model()
+
+            # --- Transcript capture via Stop hook ---
+            captured_transcript = CapturedTranscript()
+
+            def _on_stop(transcript_path: str, sdk_session_id: str) -> None:
+                captured_transcript.path = transcript_path
+                captured_transcript.sdk_session_id = sdk_session_id
+
+            security_hooks = create_security_hooks(
+                user_id,
+                sdk_cwd=sdk_cwd,
+                max_subtasks=config.claude_agent_max_subtasks,
+                on_stop=_on_stop if config.claude_agent_use_resume else None,
+            )
+
+            # --- Resume strategy: download transcript from bucket ---
+            resume_file: str | None = None
+            use_resume = False
+
+            if config.claude_agent_use_resume and user_id and len(session.messages) > 1:
+                transcript_content = await download_transcript(user_id, session_id)
+                if transcript_content and validate_transcript(transcript_content):
+                    resume_file = write_transcript_to_tempfile(
+                        transcript_content, session_id, sdk_cwd
+                    )
+                    if resume_file:
+                        use_resume = True
+                        logger.info(
+                            f"[SDK] Using --resume with transcript "
+                            f"({len(transcript_content)} bytes)"
+                        )
+
+            sdk_options_kwargs: dict[str, Any] = {
+                "system_prompt": system_prompt,
+                "mcp_servers": {"copilot": mcp_server},
+                "allowed_tools": COPILOT_TOOL_NAMES,
+                "disallowed_tools": ["Bash"],
+                "hooks": security_hooks,
+                "cwd": sdk_cwd,
+                "max_buffer_size": config.claude_agent_max_buffer_size,
+            }
+            if sdk_env:
+                sdk_options_kwargs["model"] = sdk_model
+                sdk_options_kwargs["env"] = sdk_env
+            if use_resume and resume_file:
+                sdk_options_kwargs["resume"] = resume_file
+
+            options = ClaudeAgentOptions(**sdk_options_kwargs)  # type: ignore[arg-type]
+
+            adapter = SDKResponseAdapter(message_id=message_id)
+            adapter.set_task_id(task_id)
+
+            async with ClaudeSDKClient(options=options) as client:
+                current_message = message or ""
+                if not current_message and session.messages:
+                    last_user = [m for m in session.messages if m.role == "user"]
+                    if last_user:
+                        current_message = last_user[-1].content or ""
+
+                if not current_message.strip():
+                    yield StreamError(
+                        errorText="Message cannot be empty.",
+                        code="empty_prompt",
+                    )
+                    yield StreamFinish()
+                    return
+
+                # Build query: with --resume the CLI already has full
+                # context, so we only send the new message.  Without
+                # resume, compress history into a context prefix.
+                query_message = current_message
+                if not use_resume and len(session.messages) > 1:
+                    logger.warning(
+                        f"[SDK] Using compression fallback for session "
+                        f"{session_id} ({len(session.messages)} messages) — "
+                        f"no transcript available for --resume"
+                    )
+                    compressed = await _compress_conversation_history(session)
+                    history_context = _format_conversation_context(compressed)
+                    if history_context:
+                        query_message = (
+                            f"{history_context}\n\n"
+                            f"Now, the user says:\n{current_message}"
+                        )
+
+                logger.info(
+                    f"[SDK] Sending query ({len(session.messages)} msgs in session)"
+                )
+                logger.debug(f"[SDK] Query preview: {current_message[:80]!r}")
+                await client.query(query_message, session_id=session_id)
+
+                assistant_response = ChatMessage(role="assistant", content="")
+                accumulated_tool_calls: list[dict[str, Any]] = []
+                has_appended_assistant = False
+                has_tool_results = False
+
+                async for sdk_msg in client.receive_messages():
+                    logger.debug(
+                        f"[SDK] Received: {type(sdk_msg).__name__} "
+                        f"{getattr(sdk_msg, 'subtype', '')}"
+                    )
+                    for response in adapter.convert_message(sdk_msg):
+                        if isinstance(response, StreamStart):
+                            continue
+
+                        yield response
+
+                        if isinstance(response, StreamTextDelta):
+                            delta = response.delta or ""
+                            # After tool results, start a new assistant
+                            # message for the post-tool text.
+                            if has_tool_results and has_appended_assistant:
+                                assistant_response = ChatMessage(
+                                    role="assistant", content=delta
+                                )
+                                accumulated_tool_calls = []
+                                has_appended_assistant = False
+                                has_tool_results = False
+                                session.messages.append(assistant_response)
+                                has_appended_assistant = True
+                            else:
+                                assistant_response.content = (
+                                    assistant_response.content or ""
+                                ) + delta
+                                if not has_appended_assistant:
+                                    session.messages.append(assistant_response)
+                                    has_appended_assistant = True
+
+                        elif isinstance(response, StreamToolInputAvailable):
+                            accumulated_tool_calls.append(
+                                {
+                                    "id": response.toolCallId,
+                                    "type": "function",
+                                    "function": {
+                                        "name": response.toolName,
+                                        "arguments": json.dumps(response.input or {}),
+                                    },
+                                }
+                            )
+                            assistant_response.tool_calls = accumulated_tool_calls
+                            if not has_appended_assistant:
+                                session.messages.append(assistant_response)
+                                has_appended_assistant = True
+
+                        elif isinstance(response, StreamToolOutputAvailable):
+                            session.messages.append(
+                                ChatMessage(
+                                    role="tool",
+                                    content=(
+                                        response.output
+                                        if isinstance(response.output, str)
+                                        else str(response.output)
+                                    ),
+                                    tool_call_id=response.toolCallId,
+                                )
+                            )
+                            has_tool_results = True
+
+                        elif isinstance(response, StreamFinish):
+                            stream_completed = True
+
+                    if stream_completed:
+                        break
+
+                if (
+                    assistant_response.content or assistant_response.tool_calls
+                ) and not has_appended_assistant:
+                    session.messages.append(assistant_response)
+
+                # --- Capture transcript while CLI is still alive ---
+                # Must happen INSIDE async with: close() sends SIGTERM
+                # which kills the CLI before it can flush the JSONL.
+                if (
+                    config.claude_agent_use_resume
+                    and user_id
+                    and captured_transcript.available
+                ):
+                    # Give CLI time to flush JSONL writes before we read
+                    await asyncio.sleep(0.5)
+                    raw_transcript = read_transcript_file(captured_transcript.path)
+                    if raw_transcript:
+                        task = asyncio.create_task(
+                            _upload_transcript_bg(user_id, session_id, raw_transcript)
+                        )
+                        _background_tasks.add(task)
+                        task.add_done_callback(_background_tasks.discard)
+                    else:
+                        logger.debug("[SDK] Stop hook fired but transcript not usable")
+
+        except ImportError:
+            raise RuntimeError(
+                "claude-agent-sdk is not installed. "
+                "Disable SDK mode (CHAT_USE_CLAUDE_AGENT_SDK=false) "
+                "to use the OpenAI-compatible fallback."
+            )
+
+        await upsert_chat_session(session)
+        logger.debug(
+            f"[SDK] Session {session_id} saved with {len(session.messages)} messages"
+        )
+        if not stream_completed:
+            yield StreamFinish()
+
+    except Exception as e:
+        logger.error(f"[SDK] Error: {e}", exc_info=True)
+        try:
+            await upsert_chat_session(session)
+        except Exception as save_err:
+            logger.error(f"[SDK] Failed to save session on error: {save_err}")
+        yield StreamError(
+            errorText="An error occurred. Please try again.",
+            code="sdk_error",
+        )
+        yield StreamFinish()
+    finally:
+        if sdk_cwd:
+            _cleanup_sdk_tool_results(sdk_cwd)
+
+
+async def _upload_transcript_bg(
+    user_id: str, session_id: str, raw_content: str
+) -> None:
+    """Background task to strip progress entries and upload transcript."""
+    try:
+        await upload_transcript(user_id, session_id, raw_content)
+    except Exception as e:
+        logger.error(f"[SDK] Failed to upload transcript for {session_id}: {e}")
+
+
+async def _update_title_async(
+    session_id: str, message: str, user_id: str | None = None
+) -> None:
+    """Background task to update session title."""
+    try:
+        title = await _generate_session_title(
+            message, user_id=user_id, session_id=session_id
+        )
+        if title:
+            await update_session_title(session_id, title)
+            logger.debug(f"[SDK] Generated title for {session_id}: {title}")
+    except Exception as e:
+        logger.warning(f"[SDK] Failed to update session title: {e}")

autogpt_platform/backend/backend/api/features/chat/sdk/tool_adapter.py

@@ -0,0 +1,322 @@
+"""Tool adapter for wrapping existing CoPilot tools as Claude Agent SDK MCP tools.
+
+This module provides the adapter layer that converts existing BaseTool implementations
+into in-process MCP tools that can be used with the Claude Agent SDK.
+
+Long-running tools (``is_long_running=True``) are delegated to the non-SDK
+background infrastructure (stream_registry, Redis persistence, SSE reconnection)
+via a callback provided by the service layer.  This avoids wasteful SDK polling
+and makes results survive page refreshes.
+"""
+
+import itertools
+import json
+import logging
+import os
+import uuid
+from collections.abc import Awaitable, Callable
+from contextvars import ContextVar
+from typing import Any
+
+from backend.api.features.chat.model import ChatSession
+from backend.api.features.chat.tools import TOOL_REGISTRY
+from backend.api.features.chat.tools.base import BaseTool
+
+logger = logging.getLogger(__name__)
+
+# Allowed base directory for the Read tool (SDK saves oversized tool results here).
+# Restricted to ~/.claude/projects/ and further validated to require "tool-results"
+# in the path — prevents reading settings, credentials, or other sensitive files.
+_SDK_PROJECTS_DIR = os.path.expanduser("~/.claude/projects/")
+
+# MCP server naming - the SDK prefixes tool names as "mcp__{server_name}__{tool}"
+MCP_SERVER_NAME = "copilot"
+MCP_TOOL_PREFIX = f"mcp__{MCP_SERVER_NAME}__"
+
+# Context variables to pass user/session info to tool execution
+_current_user_id: ContextVar[str | None] = ContextVar("current_user_id", default=None)
+_current_session: ContextVar[ChatSession | None] = ContextVar(
+    "current_session", default=None
+)
+# Stash for MCP tool outputs before the SDK potentially truncates them.
+# Keyed by tool_name → full output string. Consumed (popped) by the
+# response adapter when it builds StreamToolOutputAvailable.
+_pending_tool_outputs: ContextVar[dict[str, str]] = ContextVar(
+    "pending_tool_outputs", default=None  # type: ignore[arg-type]
+)
+
+# Callback type for delegating long-running tools to the non-SDK infrastructure.
+# Args: (tool_name, arguments, session) → MCP-formatted response dict.
+LongRunningCallback = Callable[
+    [str, dict[str, Any], ChatSession], Awaitable[dict[str, Any]]
+]
+
+# ContextVar so the service layer can inject the callback per-request.
+_long_running_callback: ContextVar[LongRunningCallback | None] = ContextVar(
+    "long_running_callback", default=None
+)
+
+
+def set_execution_context(
+    user_id: str | None,
+    session: ChatSession,
+    long_running_callback: LongRunningCallback | None = None,
+) -> None:
+    """Set the execution context for tool calls.
+
+    This must be called before streaming begins to ensure tools have access
+    to user_id and session information.
+
+    Args:
+        user_id: Current user's ID.
+        session: Current chat session.
+        long_running_callback: Optional callback to delegate long-running tools
+            to the non-SDK background infrastructure (stream_registry + Redis).
+    """
+    _current_user_id.set(user_id)
+    _current_session.set(session)
+    _pending_tool_outputs.set({})
+    _long_running_callback.set(long_running_callback)
+
+
+def get_execution_context() -> tuple[str | None, ChatSession | None]:
+    """Get the current execution context."""
+    return (
+        _current_user_id.get(),
+        _current_session.get(),
+    )
+
+
+def pop_pending_tool_output(tool_name: str) -> str | None:
+    """Pop and return the stashed full output for *tool_name*.
+
+    The SDK CLI may truncate large tool results (writing them to disk and
+    replacing the content with a file reference). This stash keeps the
+    original MCP output so the response adapter can forward it to the
+    frontend for proper widget rendering.
+
+    Returns ``None`` if nothing was stashed for *tool_name*.
+    """
+    pending = _pending_tool_outputs.get(None)
+    if pending is None:
+        return None
+    return pending.pop(tool_name, None)
+
+
+async def _execute_tool_sync(
+    base_tool: BaseTool,
+    user_id: str | None,
+    session: ChatSession,
+    args: dict[str, Any],
+) -> dict[str, Any]:
+    """Execute a tool synchronously and return MCP-formatted response."""
+    effective_id = f"sdk-{uuid.uuid4().hex[:12]}"
+    result = await base_tool.execute(
+        user_id=user_id,
+        session=session,
+        tool_call_id=effective_id,
+        **args,
+    )
+
+    text = (
+        result.output if isinstance(result.output, str) else json.dumps(result.output)
+    )
+
+    # Stash the full output before the SDK potentially truncates it.
+    pending = _pending_tool_outputs.get(None)
+    if pending is not None:
+        pending[base_tool.name] = text
+
+    return {
+        "content": [{"type": "text", "text": text}],
+        "isError": not result.success,
+    }
+
+
+def _mcp_error(message: str) -> dict[str, Any]:
+    return {
+        "content": [
+            {"type": "text", "text": json.dumps({"error": message, "type": "error"})}
+        ],
+        "isError": True,
+    }
+
+
+def create_tool_handler(base_tool: BaseTool):
+    """Create an async handler function for a BaseTool.
+
+    This wraps the existing BaseTool._execute method to be compatible
+    with the Claude Agent SDK MCP tool format.
+
+    Long-running tools (``is_long_running=True``) are delegated to the
+    non-SDK background infrastructure via a callback set in the execution
+    context.  The callback persists the operation in Redis (stream_registry)
+    so results survive page refreshes and pod restarts.
+    """
+
+    async def tool_handler(args: dict[str, Any]) -> dict[str, Any]:
+        """Execute the wrapped tool and return MCP-formatted response."""
+        user_id, session = get_execution_context()
+
+        if session is None:
+            return _mcp_error("No session context available")
+
+        # --- Long-running: delegate to non-SDK background infrastructure ---
+        if base_tool.is_long_running:
+            callback = _long_running_callback.get(None)
+            if callback:
+                try:
+                    return await callback(base_tool.name, args, session)
+                except Exception as e:
+                    logger.error(
+                        f"Long-running callback failed for {base_tool.name}: {e}",
+                        exc_info=True,
+                    )
+                    return _mcp_error(f"Failed to start {base_tool.name}: {e}")
+            # No callback — fall through to synchronous execution
+            logger.warning(
+                f"[SDK] No long-running callback for {base_tool.name}, "
+                f"executing synchronously (may block)"
+            )
+
+        # --- Normal (fast) tool: execute synchronously ---
+        try:
+            return await _execute_tool_sync(base_tool, user_id, session, args)
+        except Exception as e:
+            logger.error(f"Error executing tool {base_tool.name}: {e}", exc_info=True)
+            return _mcp_error(f"Failed to execute {base_tool.name}: {e}")
+
+    return tool_handler
+
+
+def _build_input_schema(base_tool: BaseTool) -> dict[str, Any]:
+    """Build a JSON Schema input schema for a tool."""
+    return {
+        "type": "object",
+        "properties": base_tool.parameters.get("properties", {}),
+        "required": base_tool.parameters.get("required", []),
+    }
+
+
+async def _read_file_handler(args: dict[str, Any]) -> dict[str, Any]:
+    """Read a file with optional offset/limit. Restricted to SDK working directory.
+
+    After reading, the file is deleted to prevent accumulation in long-running pods.
+    """
+    file_path = args.get("file_path", "")
+    offset = args.get("offset", 0)
+    limit = args.get("limit", 2000)
+
+    # Security: only allow reads under ~/.claude/projects/**/tool-results/
+    real_path = os.path.realpath(file_path)
+    if not real_path.startswith(_SDK_PROJECTS_DIR) or "tool-results" not in real_path:
+        return {
+            "content": [{"type": "text", "text": f"Access denied: {file_path}"}],
+            "isError": True,
+        }
+
+    try:
+        with open(real_path) as f:
+            selected = list(itertools.islice(f, offset, offset + limit))
+        content = "".join(selected)
+        # Cleanup happens in _cleanup_sdk_tool_results after session ends;
+        # don't delete here — the SDK may read in multiple chunks.
+        return {"content": [{"type": "text", "text": content}], "isError": False}
+    except FileNotFoundError:
+        return {
+            "content": [{"type": "text", "text": f"File not found: {file_path}"}],
+            "isError": True,
+        }
+    except Exception as e:
+        return {
+            "content": [{"type": "text", "text": f"Error reading file: {e}"}],
+            "isError": True,
+        }
+
+
+_READ_TOOL_NAME = "Read"
+_READ_TOOL_DESCRIPTION = (
+    "Read a file from the local filesystem. "
+    "Use offset and limit to read specific line ranges for large files."
+)
+_READ_TOOL_SCHEMA = {
+    "type": "object",
+    "properties": {
+        "file_path": {
+            "type": "string",
+            "description": "The absolute path to the file to read",
+        },
+        "offset": {
+            "type": "integer",
+            "description": "Line number to start reading from (0-indexed). Default: 0",
+        },
+        "limit": {
+            "type": "integer",
+            "description": "Number of lines to read. Default: 2000",
+        },
+    },
+    "required": ["file_path"],
+}
+
+
+# Create the MCP server configuration
+def create_copilot_mcp_server():
+    """Create an in-process MCP server configuration for CoPilot tools.
+
+    This can be passed to ClaudeAgentOptions.mcp_servers.
+
+    Note: The actual SDK MCP server creation depends on the claude-agent-sdk
+    package being available. This function returns the configuration that
+    can be used with the SDK.
+    """
+    try:
+        from claude_agent_sdk import create_sdk_mcp_server, tool
+
+        # Create decorated tool functions
+        sdk_tools = []
+
+        for tool_name, base_tool in TOOL_REGISTRY.items():
+            handler = create_tool_handler(base_tool)
+            decorated = tool(
+                tool_name,
+                base_tool.description,
+                _build_input_schema(base_tool),
+            )(handler)
+            sdk_tools.append(decorated)
+
+        # Add the Read tool so the SDK can read back oversized tool results
+        read_tool = tool(
+            _READ_TOOL_NAME,
+            _READ_TOOL_DESCRIPTION,
+            _READ_TOOL_SCHEMA,
+        )(_read_file_handler)
+        sdk_tools.append(read_tool)
+
+        server = create_sdk_mcp_server(
+            name=MCP_SERVER_NAME,
+            version="1.0.0",
+            tools=sdk_tools,
+        )
+
+        return server
+
+    except ImportError:
+        # Let ImportError propagate so service.py handles the fallback
+        raise
+
+
+# SDK built-in tools allowed within the workspace directory.
+# Security hooks validate that file paths stay within sdk_cwd.
+# Bash is NOT included — use the sandboxed MCP bash_exec tool instead,
+# which provides kernel-level network isolation via unshare --net.
+# Task allows spawning sub-agents (rate-limited by security hooks).
+_SDK_BUILTIN_TOOLS = ["Read", "Write", "Edit", "Glob", "Grep", "Task"]
+
+# List of tool names for allowed_tools configuration
+# Include MCP tools, the MCP Read tool for oversized results,
+# and SDK built-in file tools for workspace operations.
+COPILOT_TOOL_NAMES = [
+    *[f"{MCP_TOOL_PREFIX}{name}" for name in TOOL_REGISTRY.keys()],
+    f"{MCP_TOOL_PREFIX}{_READ_TOOL_NAME}",
+    *_SDK_BUILTIN_TOOLS,
+]

autogpt_platform/backend/backend/api/features/chat/sdk/transcript.py

@@ -0,0 +1,356 @@
+"""JSONL transcript management for stateless multi-turn resume.
+
+The Claude Code CLI persists conversations as JSONL files (one JSON object per
+line).  When the SDK's ``Stop`` hook fires we read this file, strip bloat
+(progress entries, metadata), and upload the result to bucket storage.  On the
+next turn we download the transcript, write it to a temp file, and pass
+``--resume`` so the CLI can reconstruct the full conversation.
+
+Storage is handled via ``WorkspaceStorageBackend`` (GCS in prod, local
+filesystem for self-hosted) — no DB column needed.
+"""
+
+import json
+import logging
+import os
+import re
+
+logger = logging.getLogger(__name__)
+
+# UUIDs are hex + hyphens; strip everything else to prevent path injection.
+_SAFE_ID_RE = re.compile(r"[^0-9a-fA-F-]")
+
+# Entry types that can be safely removed from the transcript without breaking
+# the parentUuid conversation tree that ``--resume`` relies on.
+# - progress: UI progress ticks, no message content (avg 97KB for agent_progress)
+# - file-history-snapshot: undo tracking metadata
+# - queue-operation: internal queue bookkeeping
+# - summary: session summaries
+# - pr-link: PR link metadata
+STRIPPABLE_TYPES = frozenset(
+    {"progress", "file-history-snapshot", "queue-operation", "summary", "pr-link"}
+)
+
+# Workspace storage constants — deterministic path from session_id.
+TRANSCRIPT_STORAGE_PREFIX = "chat-transcripts"
+
+
+# ---------------------------------------------------------------------------
+# Progress stripping
+# ---------------------------------------------------------------------------
+
+
+def strip_progress_entries(content: str) -> str:
+    """Remove progress/metadata entries from a JSONL transcript.
+
+    Removes entries whose ``type`` is in ``STRIPPABLE_TYPES`` and reparents
+    any remaining child entries so the ``parentUuid`` chain stays intact.
+    Typically reduces transcript size by ~30%.
+    """
+    lines = content.strip().split("\n")
+
+    entries: list[dict] = []
+    for line in lines:
+        try:
+            entries.append(json.loads(line))
+        except json.JSONDecodeError:
+            # Keep unparseable lines as-is (safety)
+            entries.append({"_raw": line})
+
+    stripped_uuids: set[str] = set()
+    uuid_to_parent: dict[str, str] = {}
+    kept: list[dict] = []
+
+    for entry in entries:
+        if "_raw" in entry:
+            kept.append(entry)
+            continue
+        uid = entry.get("uuid", "")
+        parent = entry.get("parentUuid", "")
+        entry_type = entry.get("type", "")
+
+        if uid:
+            uuid_to_parent[uid] = parent
+
+        if entry_type in STRIPPABLE_TYPES:
+            if uid:
+                stripped_uuids.add(uid)
+        else:
+            kept.append(entry)
+
+    # Reparent: walk up chain through stripped entries to find surviving ancestor
+    for entry in kept:
+        if "_raw" in entry:
+            continue
+        parent = entry.get("parentUuid", "")
+        original_parent = parent
+        while parent in stripped_uuids:
+            parent = uuid_to_parent.get(parent, "")
+        if parent != original_parent:
+            entry["parentUuid"] = parent
+
+    result_lines: list[str] = []
+    for entry in kept:
+        if "_raw" in entry:
+            result_lines.append(entry["_raw"])
+        else:
+            result_lines.append(json.dumps(entry, separators=(",", ":")))
+
+    return "\n".join(result_lines) + "\n"
+
+
+# ---------------------------------------------------------------------------
+# Local file I/O (read from CLI's JSONL, write temp file for --resume)
+# ---------------------------------------------------------------------------
+
+
+def read_transcript_file(transcript_path: str) -> str | None:
+    """Read a JSONL transcript file from disk.
+
+    Returns the raw JSONL content, or ``None`` if the file is missing, empty,
+    or only contains metadata (≤2 lines with no conversation messages).
+    """
+    if not transcript_path or not os.path.isfile(transcript_path):
+        logger.debug(f"[Transcript] File not found: {transcript_path}")
+        return None
+
+    try:
+        with open(transcript_path) as f:
+            content = f.read()
+
+        if not content.strip():
+            logger.debug(f"[Transcript] Empty file: {transcript_path}")
+            return None
+
+        lines = content.strip().split("\n")
+        if len(lines) < 3:
+            # Raw files with ≤2 lines are metadata-only
+            # (queue-operation + file-history-snapshot, no conversation).
+            logger.debug(
+                f"[Transcript] Too few lines ({len(lines)}): {transcript_path}"
+            )
+            return None
+
+        # Quick structural validation — parse first and last lines.
+        json.loads(lines[0])
+        json.loads(lines[-1])
+
+        logger.info(
+            f"[Transcript] Read {len(lines)} lines, "
+            f"{len(content)} bytes from {transcript_path}"
+        )
+        return content
+
+    except (json.JSONDecodeError, OSError) as e:
+        logger.warning(f"[Transcript] Failed to read {transcript_path}: {e}")
+        return None
+
+
+def _sanitize_id(raw_id: str, max_len: int = 36) -> str:
+    """Sanitize an ID for safe use in file paths.
+
+    Session/user IDs are expected to be UUIDs (hex + hyphens).  Strip
+    everything else and truncate to *max_len* so the result cannot introduce
+    path separators or other special characters.
+    """
+    cleaned = _SAFE_ID_RE.sub("", raw_id or "")[:max_len]
+    return cleaned or "unknown"
+
+
+_SAFE_CWD_PREFIX = os.path.realpath("/tmp/copilot-")
+
+
+def write_transcript_to_tempfile(
+    transcript_content: str,
+    session_id: str,
+    cwd: str,
+) -> str | None:
+    """Write JSONL transcript to a temp file inside *cwd* for ``--resume``.
+
+    The file lives in the session working directory so it is cleaned up
+    automatically when the session ends.
+
+    Returns the absolute path to the file, or ``None`` on failure.
+    """
+    # Validate cwd is under the expected sandbox prefix (CodeQL sanitizer).
+    real_cwd = os.path.realpath(cwd)
+    if not real_cwd.startswith(_SAFE_CWD_PREFIX):
+        logger.warning(f"[Transcript] cwd outside sandbox: {cwd}")
+        return None
+
+    try:
+        os.makedirs(real_cwd, exist_ok=True)
+        safe_id = _sanitize_id(session_id, max_len=8)
+        jsonl_path = os.path.realpath(
+            os.path.join(real_cwd, f"transcript-{safe_id}.jsonl")
+        )
+        if not jsonl_path.startswith(real_cwd):
+            logger.warning(f"[Transcript] Path escaped cwd: {jsonl_path}")
+            return None
+
+        with open(jsonl_path, "w") as f:
+            f.write(transcript_content)
+
+        logger.info(f"[Transcript] Wrote resume file: {jsonl_path}")
+        return jsonl_path
+
+    except OSError as e:
+        logger.warning(f"[Transcript] Failed to write resume file: {e}")
+        return None
+
+
+def validate_transcript(content: str | None) -> bool:
+    """Check that a transcript has actual conversation messages.
+
+    A valid transcript for resume needs at least one user message and one
+    assistant message (not just queue-operation / file-history-snapshot
+    metadata).
+    """
+    if not content or not content.strip():
+        return False
+
+    lines = content.strip().split("\n")
+    if len(lines) < 2:
+        return False
+
+    has_user = False
+    has_assistant = False
+
+    for line in lines:
+        try:
+            entry = json.loads(line)
+            msg_type = entry.get("type")
+            if msg_type == "user":
+                has_user = True
+            elif msg_type == "assistant":
+                has_assistant = True
+        except json.JSONDecodeError:
+            return False
+
+    return has_user and has_assistant
+
+
+# ---------------------------------------------------------------------------
+# Bucket storage (GCS / local via WorkspaceStorageBackend)
+# ---------------------------------------------------------------------------
+
+
+def _storage_path_parts(user_id: str, session_id: str) -> tuple[str, str, str]:
+    """Return (workspace_id, file_id, filename) for a session's transcript.
+
+    Path structure: ``chat-transcripts/{user_id}/{session_id}.jsonl``
+    IDs are sanitized to hex+hyphen to prevent path traversal.
+    """
+    return (
+        TRANSCRIPT_STORAGE_PREFIX,
+        _sanitize_id(user_id),
+        f"{_sanitize_id(session_id)}.jsonl",
+    )
+
+
+def _build_storage_path(user_id: str, session_id: str, backend: object) -> str:
+    """Build the full storage path string that ``retrieve()`` expects.
+
+    ``store()`` returns a path like ``gcs://bucket/workspaces/...`` or
+    ``local://workspace_id/file_id/filename``.  Since we use deterministic
+    arguments we can reconstruct the same path for download/delete without
+    having stored the return value.
+    """
+    from backend.util.workspace_storage import GCSWorkspaceStorage
+
+    wid, fid, fname = _storage_path_parts(user_id, session_id)
+
+    if isinstance(backend, GCSWorkspaceStorage):
+        blob = f"workspaces/{wid}/{fid}/{fname}"
+        return f"gcs://{backend.bucket_name}/{blob}"
+    else:
+        # LocalWorkspaceStorage returns local://{relative_path}
+        return f"local://{wid}/{fid}/{fname}"
+
+
+async def upload_transcript(user_id: str, session_id: str, content: str) -> None:
+    """Strip progress entries and upload transcript to bucket storage.
+
+    Safety: only overwrites when the new (stripped) transcript is larger than
+    what is already stored.  Since JSONL is append-only, the latest transcript
+    is always the longest.  This prevents a slow/stale background task from
+    clobbering a newer upload from a concurrent turn.
+    """
+    from backend.util.workspace_storage import get_workspace_storage
+
+    stripped = strip_progress_entries(content)
+    if not validate_transcript(stripped):
+        logger.warning(
+            f"[Transcript] Skipping upload — stripped content is not a valid "
+            f"transcript for session {session_id}"
+        )
+        return
+
+    storage = await get_workspace_storage()
+    wid, fid, fname = _storage_path_parts(user_id, session_id)
+    encoded = stripped.encode("utf-8")
+    new_size = len(encoded)
+
+    # Check existing transcript size to avoid overwriting newer with older
+    path = _build_storage_path(user_id, session_id, storage)
+    try:
+        existing = await storage.retrieve(path)
+        if len(existing) >= new_size:
+            logger.info(
+                f"[Transcript] Skipping upload — existing transcript "
+                f"({len(existing)}B) >= new ({new_size}B) for session "
+                f"{session_id}"
+            )
+            return
+    except (FileNotFoundError, Exception):
+        pass  # No existing transcript or retrieval error — proceed with upload
+
+    await storage.store(
+        workspace_id=wid,
+        file_id=fid,
+        filename=fname,
+        content=encoded,
+    )
+    logger.info(
+        f"[Transcript] Uploaded {new_size} bytes "
+        f"(stripped from {len(content)}) for session {session_id}"
+    )
+
+
+async def download_transcript(user_id: str, session_id: str) -> str | None:
+    """Download transcript from bucket storage.
+
+    Returns the JSONL content string, or ``None`` if not found.
+    """
+    from backend.util.workspace_storage import get_workspace_storage
+
+    storage = await get_workspace_storage()
+    path = _build_storage_path(user_id, session_id, storage)
+
+    try:
+        data = await storage.retrieve(path)
+        content = data.decode("utf-8")
+        logger.info(
+            f"[Transcript] Downloaded {len(content)} bytes for session {session_id}"
+        )
+        return content
+    except FileNotFoundError:
+        logger.debug(f"[Transcript] No transcript in storage for {session_id}")
+        return None
+    except Exception as e:
+        logger.warning(f"[Transcript] Failed to download transcript: {e}")
+        return None
+
+
+async def delete_transcript(user_id: str, session_id: str) -> None:
+    """Delete transcript from bucket storage (e.g. after resume failure)."""
+    from backend.util.workspace_storage import get_workspace_storage
+
+    storage = await get_workspace_storage()
+    path = _build_storage_path(user_id, session_id, storage)
+
+    try:
+        await storage.delete(path)
+        logger.info(f"[Transcript] Deleted transcript for session {session_id}")
+    except Exception as e:
+        logger.warning(f"[Transcript] Failed to delete transcript: {e}")

autogpt_platform/backend/backend/api/features/chat/service.py

@@ -245,12 +245,16 @@ async def _get_system_prompt_template(context: str) -> str:
     return DEFAULT_SYSTEM_PROMPT.format(users_information=context)
 
 
-async def _build_system_prompt(user_id: str | None) -> tuple[str, Any]:
+async def _build_system_prompt(
+    user_id: str | None, has_conversation_history: bool = False
+) -> tuple[str, Any]:
     """Build the full system prompt including business understanding if available.
 
     Args:
-        user_id: The user ID for fetching business understanding
-                     If "default" and this is the user's first session, will use "onboarding" instead.
+        user_id: The user ID for fetching business understanding.
+        has_conversation_history: Whether there's existing conversation history.
+            If True, we don't tell the model to greet/introduce (since they're
+            already in a conversation).
 
     Returns:
         Tuple of (compiled prompt string, business understanding object)
@@ -266,6 +270,8 @@ async def _build_system_prompt(user_id: str | None) -> tuple[str, Any]:
 
     if understanding:
         context = format_understanding_for_prompt(understanding)
+    elif has_conversation_history:
+        context = "No prior understanding saved yet. Continue the existing conversation naturally."
     else:
         context = "This is the first time you are meeting the user. Greet them and introduce them to the platform"
 
@@ -374,7 +380,6 @@ async def stream_chat_completion(
 
     Raises:
         NotFoundError: If session_id is invalid
-        ValueError: If max_context_messages is exceeded
 
     """
     completion_start = time.monotonic()
@@ -459,8 +464,9 @@ async def stream_chat_completion(
 
     # Generate title for new sessions on first user message (non-blocking)
     # Check: is_user_message, no title yet, and this is the first user message
-    if is_user_message and message and not session.title:
-        user_messages = [m for m in session.messages if m.role == "user"]
+    user_messages = [m for m in session.messages if m.role == "user"]
+    first_user_msg = message or (user_messages[0].content if user_messages else None)
+    if is_user_message and first_user_msg and not session.title:
         if len(user_messages) == 1:
             # First user message - generate title in background
             import asyncio
@@ -468,7 +474,7 @@ async def stream_chat_completion(
             # Capture only the values we need (not the session object) to avoid
             # stale data issues when the main flow modifies the session
             captured_session_id = session_id
-            captured_message = message
+            captured_message = first_user_msg
             captured_user_id = user_id
 
             async def _update_title():
@@ -1237,7 +1243,7 @@ async def _stream_chat_chunks(
 
                 total_time = (time_module.perf_counter() - stream_chunks_start) * 1000
                 logger.info(
-                    f"[TIMING] _stream_chat_chunks COMPLETED in {total_time/1000:.1f}s; "
+                    f"[TIMING] _stream_chat_chunks COMPLETED in {total_time / 1000:.1f}s; "
                     f"session={session.session_id}, user={session.user_id}",
                     extra={"json_fields": {**log_meta, "total_time_ms": total_time}},
                 )
@@ -1245,6 +1251,7 @@ async def _stream_chat_chunks(
                 return
             except Exception as e:
                 last_error = e
+
                 if _is_retryable_error(e) and retry_count < MAX_RETRIES:
                     retry_count += 1
                     # Calculate delay with exponential backoff
@@ -1260,12 +1267,27 @@ async def _stream_chat_chunks(
                     continue  # Retry the stream
                 else:
                     # Non-retryable error or max retries exceeded
-                    logger.error(
-                        f"Error in stream (not retrying): {e!s}",
-                        exc_info=True,
+                    _log_api_error(
+                        error=e,
+                        context="stream (not retrying)",
+                        session_id=session.session_id if session else None,
+                        message_count=len(messages) if messages else None,
+                        model=model,
+                        retry_count=retry_count,
                     )
                     error_code = None
                     error_text = str(e)
+
+                    error_details = _extract_api_error_details(e)
+                    if error_details.get("response_body"):
+                        body = error_details["response_body"]
+                        if isinstance(body, dict):
+                            err = body.get("error")
+                            if isinstance(err, dict) and err.get("message"):
+                                error_text = err["message"]
+                            elif body.get("message"):
+                                error_text = body["message"]
+
                     if _is_region_blocked_error(e):
                         error_code = "MODEL_NOT_AVAILABLE_REGION"
                         error_text = (
@@ -1282,9 +1304,13 @@ async def _stream_chat_chunks(
 
         # If we exit the retry loop without returning, it means we exhausted retries
         if last_error:
-            logger.error(
-                f"Max retries ({MAX_RETRIES}) exceeded. Last error: {last_error!s}",
-                exc_info=True,
+            _log_api_error(
+                error=last_error,
+                context=f"stream (max retries {MAX_RETRIES} exceeded)",
+                session_id=session.session_id if session else None,
+                message_count=len(messages) if messages else None,
+                model=model,
+                retry_count=MAX_RETRIES,
             )
             yield StreamError(errorText=f"Max retries exceeded: {last_error!s}")
             yield StreamFinish()
@@ -1857,6 +1883,7 @@ async def _generate_llm_continuation(
                 break  # Success, exit retry loop
             except Exception as e:
                 last_error = e
+
                 if _is_retryable_error(e) and retry_count < MAX_RETRIES:
                     retry_count += 1
                     delay = min(
@@ -1870,17 +1897,25 @@ async def _generate_llm_continuation(
                     await asyncio.sleep(delay)
                     continue
                 else:
-                    # Non-retryable error - log and exit gracefully
-                    logger.error(
-                        f"Non-retryable error in LLM continuation: {e!s}",
-                        exc_info=True,
+                    # Non-retryable error - log details and exit gracefully
+                    _log_api_error(
+                        error=e,
+                        context="LLM continuation (not retrying)",
+                        session_id=session_id,
+                        message_count=len(messages) if messages else None,
+                        model=config.model,
+                        retry_count=retry_count,
                     )
                     return
 
         if last_error:
-            logger.error(
-                f"Max retries ({MAX_RETRIES}) exceeded for LLM continuation. "
-                f"Last error: {last_error!s}"
+            _log_api_error(
+                error=last_error,
+                context=f"LLM continuation (max retries {MAX_RETRIES} exceeded)",
+                session_id=session_id,
+                message_count=len(messages) if messages else None,
+                model=config.model,
+                retry_count=MAX_RETRIES,
             )
             return
 
@@ -1920,6 +1955,91 @@ async def _generate_llm_continuation(
         logger.error(f"Failed to generate LLM continuation: {e}", exc_info=True)
 
 
+def _log_api_error(
+    error: Exception,
+    context: str,
+    session_id: str | None = None,
+    message_count: int | None = None,
+    model: str | None = None,
+    retry_count: int = 0,
+) -> None:
+    """Log detailed API error information for debugging."""
+    details = _extract_api_error_details(error)
+    details["context"] = context
+    details["session_id"] = session_id
+    details["message_count"] = message_count
+    details["model"] = model
+    details["retry_count"] = retry_count
+
+    if isinstance(error, RateLimitError):
+        logger.warning(f"Rate limit error in {context}: {details}", exc_info=error)
+    elif isinstance(error, APIConnectionError):
+        logger.warning(f"API connection error in {context}: {details}", exc_info=error)
+    elif isinstance(error, APIStatusError) and error.status_code >= 500:
+        logger.error(f"API server error (5xx) in {context}: {details}", exc_info=error)
+    else:
+        logger.error(f"API error in {context}: {details}", exc_info=error)
+
+
+def _extract_api_error_details(error: Exception) -> dict[str, Any]:
+    """Extract detailed information from OpenAI/OpenRouter API errors."""
+    error_msg = str(error)
+    details: dict[str, Any] = {
+        "error_type": type(error).__name__,
+        "error_message": error_msg[:500] + "..." if len(error_msg) > 500 else error_msg,
+    }
+
+    if hasattr(error, "code"):
+        details["code"] = getattr(error, "code", None)
+    if hasattr(error, "param"):
+        details["param"] = getattr(error, "param", None)
+
+    if isinstance(error, APIStatusError):
+        details["status_code"] = error.status_code
+        details["request_id"] = getattr(error, "request_id", None)
+
+        if hasattr(error, "body") and error.body:
+            details["response_body"] = _sanitize_error_body(error.body)
+
+        if hasattr(error, "response") and error.response:
+            headers = error.response.headers
+            details["openrouter_provider"] = headers.get("x-openrouter-provider")
+            details["openrouter_model"] = headers.get("x-openrouter-model")
+            details["retry_after"] = headers.get("retry-after")
+            details["rate_limit_remaining"] = headers.get("x-ratelimit-remaining")
+
+    return details
+
+
+def _sanitize_error_body(
+    body: Any, max_length: int = 2000
+) -> dict[str, Any] | str | None:
+    """Extract only safe fields from error response body to avoid logging sensitive data."""
+    if not isinstance(body, dict):
+        # Non-dict bodies (e.g., HTML error pages) - return truncated string
+        if body is not None:
+            body_str = str(body)
+            if len(body_str) > max_length:
+                return body_str[:max_length] + "...[truncated]"
+            return body_str
+        return None
+
+    safe_fields = ("message", "type", "code", "param", "error")
+    sanitized: dict[str, Any] = {}
+
+    for field in safe_fields:
+        if field in body:
+            value = body[field]
+            if field == "error" and isinstance(value, dict):
+                sanitized[field] = _sanitize_error_body(value, max_length)
+            elif isinstance(value, str) and len(value) > max_length:
+                sanitized[field] = value[:max_length] + "...[truncated]"
+            else:
+                sanitized[field] = value
+
+    return sanitized if sanitized else None
+
+
 async def _generate_llm_continuation_with_streaming(
     session_id: str,
     user_id: str | None,

autogpt_platform/backend/backend/api/features/chat/service_test.py

@@ -1,3 +1,4 @@
+import asyncio
 import logging
 from os import getenv
 
@@ -11,6 +12,8 @@ from .response_model import (
     StreamTextDelta,
     StreamToolOutputAvailable,
 )
+from .sdk import service as sdk_service
+from .sdk.transcript import download_transcript
 
 logger = logging.getLogger(__name__)
 
@@ -80,3 +83,96 @@ async def test_stream_chat_completion_with_tool_calls(setup_test_user, test_user
     session = await get_chat_session(session.session_id)
     assert session, "Session not found"
     assert session.usage, "Usage is empty"
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_sdk_resume_multi_turn(setup_test_user, test_user_id):
+    """Test that the SDK --resume path captures and uses transcripts across turns.
+
+    Turn 1: Send a message containing a unique keyword.
+    Turn 2: Ask the model to recall that keyword — proving the transcript was
+    persisted and restored via --resume.
+    """
+    api_key: str | None = getenv("OPEN_ROUTER_API_KEY")
+    if not api_key:
+        return pytest.skip("OPEN_ROUTER_API_KEY is not set, skipping test")
+
+    from .config import ChatConfig
+
+    cfg = ChatConfig()
+    if not cfg.claude_agent_use_resume:
+        return pytest.skip("CLAUDE_AGENT_USE_RESUME is not enabled, skipping test")
+
+    session = await create_chat_session(test_user_id)
+    session = await upsert_chat_session(session)
+
+    # --- Turn 1: send a message with a unique keyword ---
+    keyword = "ZEPHYR42"
+    turn1_msg = (
+        f"Please remember this special keyword: {keyword}. "
+        "Just confirm you've noted it, keep your response brief."
+    )
+    turn1_text = ""
+    turn1_errors: list[str] = []
+    turn1_ended = False
+
+    async for chunk in sdk_service.stream_chat_completion_sdk(
+        session.session_id,
+        turn1_msg,
+        user_id=test_user_id,
+    ):
+        if isinstance(chunk, StreamTextDelta):
+            turn1_text += chunk.delta
+        elif isinstance(chunk, StreamError):
+            turn1_errors.append(chunk.errorText)
+        elif isinstance(chunk, StreamFinish):
+            turn1_ended = True
+
+    assert turn1_ended, "Turn 1 did not finish"
+    assert not turn1_errors, f"Turn 1 errors: {turn1_errors}"
+    assert turn1_text, "Turn 1 produced no text"
+
+    # Wait for background upload task to complete (retry up to 5s)
+    transcript = None
+    for _ in range(10):
+        await asyncio.sleep(0.5)
+        transcript = await download_transcript(test_user_id, session.session_id)
+        if transcript:
+            break
+    assert transcript, (
+        "Transcript was not uploaded to bucket after turn 1 — "
+        "Stop hook may not have fired or transcript was too small"
+    )
+    logger.info(f"Turn 1 transcript uploaded: {len(transcript)} bytes")
+
+    # Reload session for turn 2
+    session = await get_chat_session(session.session_id, test_user_id)
+    assert session, "Session not found after turn 1"
+
+    # --- Turn 2: ask model to recall the keyword ---
+    turn2_msg = "What was the special keyword I asked you to remember?"
+    turn2_text = ""
+    turn2_errors: list[str] = []
+    turn2_ended = False
+
+    async for chunk in sdk_service.stream_chat_completion_sdk(
+        session.session_id,
+        turn2_msg,
+        user_id=test_user_id,
+        session=session,
+    ):
+        if isinstance(chunk, StreamTextDelta):
+            turn2_text += chunk.delta
+        elif isinstance(chunk, StreamError):
+            turn2_errors.append(chunk.errorText)
+        elif isinstance(chunk, StreamFinish):
+            turn2_ended = True
+
+    assert turn2_ended, "Turn 2 did not finish"
+    assert not turn2_errors, f"Turn 2 errors: {turn2_errors}"
+    assert turn2_text, "Turn 2 produced no text"
+    assert keyword in turn2_text, (
+        f"Model did not recall keyword '{keyword}' in turn 2. "
+        f"Response: {turn2_text[:200]}"
+    )
+    logger.info(f"Turn 2 recalled keyword successfully: {turn2_text[:100]}")

autogpt_platform/backend/backend/api/features/chat/stream_registry.py

@@ -814,6 +814,28 @@ async def get_active_task_for_session(
                 if task_user_id and user_id != task_user_id:
                     continue
 
+                # Auto-expire stale tasks that exceeded stream_timeout
+                created_at_str = meta.get("created_at", "")
+                if created_at_str:
+                    try:
+                        created_at = datetime.fromisoformat(created_at_str)
+                        age_seconds = (
+                            datetime.now(timezone.utc) - created_at
+                        ).total_seconds()
+                        if age_seconds > config.stream_timeout:
+                            logger.warning(
+                                f"[TASK_LOOKUP] Auto-expiring stale task {task_id[:8]}... "
+                                f"(age={age_seconds:.0f}s > timeout={config.stream_timeout}s)"
+                            )
+                            await mark_task_completed(task_id, "failed")
+                            continue
+                    except (ValueError, TypeError):
+                        pass
+
+                logger.info(
+                    f"[TASK_LOOKUP] Found running task {task_id[:8]}... for session {session_id[:8]}..."
+                )
+
                 # Get the last message ID from Redis Stream
                 stream_key = _get_task_stream_key(task_id)
                 last_id = "0-0"

autogpt_platform/backend/backend/api/features/chat/tools/__init__.py

@@ -9,9 +9,12 @@ from backend.api.features.chat.tracking import track_tool_called
 from .add_understanding import AddUnderstandingTool
 from .agent_output import AgentOutputTool
 from .base import BaseTool
+from .bash_exec import BashExecTool
+from .check_operation_status import CheckOperationStatusTool
 from .create_agent import CreateAgentTool
 from .customize_agent import CustomizeAgentTool
 from .edit_agent import EditAgentTool
+from .feature_requests import CreateFeatureRequestTool, SearchFeatureRequestsTool
 from .find_agent import FindAgentTool
 from .find_block import FindBlockTool
 from .find_library_agent import FindLibraryAgentTool
@@ -19,6 +22,7 @@ from .get_doc_page import GetDocPageTool
 from .run_agent import RunAgentTool
 from .run_block import RunBlockTool
 from .search_docs import SearchDocsTool
+from .web_fetch import WebFetchTool
 from .workspace_files import (
     DeleteWorkspaceFileTool,
     ListWorkspaceFilesTool,
@@ -43,8 +47,17 @@ TOOL_REGISTRY: dict[str, BaseTool] = {
     "run_agent": RunAgentTool(),
     "run_block": RunBlockTool(),
     "view_agent_output": AgentOutputTool(),
+    "check_operation_status": CheckOperationStatusTool(),
     "search_docs": SearchDocsTool(),
     "get_doc_page": GetDocPageTool(),
+    # Web fetch for safe URL retrieval
+    "web_fetch": WebFetchTool(),
+    # Sandboxed code execution (bubblewrap)
+    "bash_exec": BashExecTool(),
+    # Persistent workspace tools (cloud storage, survives across sessions)
+    # Feature request tools
+    "search_feature_requests": SearchFeatureRequestsTool(),
+    "create_feature_request": CreateFeatureRequestTool(),
     # Workspace tools for CoPilot file operations
     "list_workspace_files": ListWorkspaceFilesTool(),
     "read_workspace_file": ReadWorkspaceFileTool(),

autogpt_platform/backend/backend/api/features/chat/tools/bash_exec.py

@@ -0,0 +1,131 @@
+"""Bash execution tool — run shell commands in a bubblewrap sandbox.
+
+Full Bash scripting is allowed (loops, conditionals, pipes, functions, etc.).
+Safety comes from OS-level isolation (bubblewrap): only system dirs visible
+read-only, writable workspace only, clean env, no network.
+
+Requires bubblewrap (``bwrap``) — the tool is disabled when bwrap is not
+available (e.g. macOS development).
+"""
+
+import logging
+from typing import Any
+
+from backend.api.features.chat.model import ChatSession
+from backend.api.features.chat.tools.base import BaseTool
+from backend.api.features.chat.tools.models import (
+    BashExecResponse,
+    ErrorResponse,
+    ToolResponseBase,
+)
+from backend.api.features.chat.tools.sandbox import (
+    get_workspace_dir,
+    has_full_sandbox,
+    run_sandboxed,
+)
+
+logger = logging.getLogger(__name__)
+
+
+class BashExecTool(BaseTool):
+    """Execute Bash commands in a bubblewrap sandbox."""
+
+    @property
+    def name(self) -> str:
+        return "bash_exec"
+
+    @property
+    def description(self) -> str:
+        if not has_full_sandbox():
+            return (
+                "Bash execution is DISABLED — bubblewrap sandbox is not "
+                "available on this platform. Do not call this tool."
+            )
+        return (
+            "Execute a Bash command or script in a bubblewrap sandbox. "
+            "Full Bash scripting is supported (loops, conditionals, pipes, "
+            "functions, etc.). "
+            "The sandbox shares the same working directory as the SDK Read/Write "
+            "tools — files created by either are accessible to both. "
+            "SECURITY: Only system directories (/usr, /bin, /lib, /etc) are "
+            "visible read-only, the per-session workspace is the only writable "
+            "path, environment variables are wiped (no secrets), all network "
+            "access is blocked at the kernel level, and resource limits are "
+            "enforced (max 64 processes, 512MB memory, 50MB file size). "
+            "Application code, configs, and other directories are NOT accessible. "
+            "To fetch web content, use the web_fetch tool instead. "
+            "Execution is killed after the timeout (default 30s, max 120s). "
+            "Returns stdout and stderr. "
+            "Useful for file manipulation, data processing with Unix tools "
+            "(grep, awk, sed, jq, etc.), and running shell scripts."
+        )
+
+    @property
+    def parameters(self) -> dict[str, Any]:
+        return {
+            "type": "object",
+            "properties": {
+                "command": {
+                    "type": "string",
+                    "description": "Bash command or script to execute.",
+                },
+                "timeout": {
+                    "type": "integer",
+                    "description": (
+                        "Max execution time in seconds (default 30, max 120)."
+                    ),
+                    "default": 30,
+                },
+            },
+            "required": ["command"],
+        }
+
+    @property
+    def requires_auth(self) -> bool:
+        return False
+
+    async def _execute(
+        self,
+        user_id: str | None,
+        session: ChatSession,
+        **kwargs: Any,
+    ) -> ToolResponseBase:
+        session_id = session.session_id if session else None
+
+        if not has_full_sandbox():
+            return ErrorResponse(
+                message="bash_exec requires bubblewrap sandbox (Linux only).",
+                error="sandbox_unavailable",
+                session_id=session_id,
+            )
+
+        command: str = (kwargs.get("command") or "").strip()
+        timeout: int = kwargs.get("timeout", 30)
+
+        if not command:
+            return ErrorResponse(
+                message="No command provided.",
+                error="empty_command",
+                session_id=session_id,
+            )
+
+        workspace = get_workspace_dir(session_id or "default")
+
+        stdout, stderr, exit_code, timed_out = await run_sandboxed(
+            command=["bash", "-c", command],
+            cwd=workspace,
+            timeout=timeout,
+        )
+
+        return BashExecResponse(
+            message=(
+                "Execution timed out"
+                if timed_out
+                else f"Command executed (exit {exit_code})"
+            ),
+            stdout=stdout,
+            stderr=stderr,
+            exit_code=exit_code,
+            timed_out=timed_out,
+            session_id=session_id,
+        )

autogpt_platform/backend/backend/api/features/chat/tools/check_operation_status.py

@@ -0,0 +1,127 @@
+"""CheckOperationStatusTool — query the status of a long-running operation."""
+
+import logging
+from typing import Any
+
+from backend.api.features.chat.model import ChatSession
+from backend.api.features.chat.tools.base import BaseTool
+from backend.api.features.chat.tools.models import (
+    ErrorResponse,
+    ResponseType,
+    ToolResponseBase,
+)
+
+logger = logging.getLogger(__name__)
+
+
+class OperationStatusResponse(ToolResponseBase):
+    """Response for check_operation_status tool."""
+
+    type: ResponseType = ResponseType.OPERATION_STATUS
+    task_id: str
+    operation_id: str
+    status: str  # "running", "completed", "failed"
+    tool_name: str | None = None
+    message: str = ""
+
+
+class CheckOperationStatusTool(BaseTool):
+    """Check the status of a long-running operation (create_agent, edit_agent, etc.).
+
+    The CoPilot uses this tool to report back to the user whether an
+    operation that was started earlier has completed, failed, or is still
+    running.
+    """
+
+    @property
+    def name(self) -> str:
+        return "check_operation_status"
+
+    @property
+    def description(self) -> str:
+        return (
+            "Check the current status of a long-running operation such as "
+            "create_agent or edit_agent. Accepts either an operation_id or "
+            "task_id from a previous operation_started response. "
+            "Returns the current status: running, completed, or failed."
+        )
+
+    @property
+    def parameters(self) -> dict[str, Any]:
+        return {
+            "type": "object",
+            "properties": {
+                "operation_id": {
+                    "type": "string",
+                    "description": (
+                        "The operation_id from an operation_started response."
+                    ),
+                },
+                "task_id": {
+                    "type": "string",
+                    "description": (
+                        "The task_id from an operation_started response. "
+                        "Used as fallback if operation_id is not provided."
+                    ),
+                },
+            },
+            "required": [],
+        }
+
+    @property
+    def requires_auth(self) -> bool:
+        return False
+
+    async def _execute(
+        self,
+        user_id: str | None,
+        session: ChatSession,
+        **kwargs,
+    ) -> ToolResponseBase:
+        from backend.api.features.chat import stream_registry
+
+        operation_id = (kwargs.get("operation_id") or "").strip()
+        task_id = (kwargs.get("task_id") or "").strip()
+
+        if not operation_id and not task_id:
+            return ErrorResponse(
+                message="Please provide an operation_id or task_id.",
+                error="missing_parameter",
+            )
+
+        task = None
+        if operation_id:
+            task = await stream_registry.find_task_by_operation_id(operation_id)
+        if task is None and task_id:
+            task = await stream_registry.get_task(task_id)
+
+        if task is None:
+            # Task not in Redis — it may have already expired (TTL).
+            # Check conversation history for the result instead.
+            return ErrorResponse(
+                message=(
+                    "Operation not found — it may have already completed and "
+                    "expired from the status tracker. Check the conversation "
+                    "history for the result."
+                ),
+                error="not_found",
+            )
+
+        status_messages = {
+            "running": (
+                f"The {task.tool_name or 'operation'} is still running. "
+                "Please wait for it to complete."
+            ),
+            "completed": (
+                f"The {task.tool_name or 'operation'} has completed successfully."
+            ),
+            "failed": f"The {task.tool_name or 'operation'} has failed.",
+        }
+
+        return OperationStatusResponse(
+            task_id=task.task_id,
+            operation_id=task.operation_id,
+            status=task.status,
+            tool_name=task.tool_name,
+            message=status_messages.get(task.status, f"Status: {task.status}"),
+        )

autogpt_platform/backend/backend/api/features/chat/tools/feature_requests.py

@@ -0,0 +1,448 @@
+"""Feature request tools - search and create feature requests via Linear."""
+
+import logging
+from typing import Any
+
+from pydantic import SecretStr
+
+from backend.api.features.chat.model import ChatSession
+from backend.api.features.chat.tools.base import BaseTool
+from backend.api.features.chat.tools.models import (
+    ErrorResponse,
+    FeatureRequestCreatedResponse,
+    FeatureRequestInfo,
+    FeatureRequestSearchResponse,
+    NoResultsResponse,
+    ToolResponseBase,
+)
+from backend.blocks.linear._api import LinearClient
+from backend.data.model import APIKeyCredentials
+from backend.data.user import get_user_email_by_id
+from backend.util.settings import Settings
+
+logger = logging.getLogger(__name__)
+
+MAX_SEARCH_RESULTS = 10
+
+# GraphQL queries/mutations
+SEARCH_ISSUES_QUERY = """
+query SearchFeatureRequests($term: String!, $filter: IssueFilter, $first: Int) {
+  searchIssues(term: $term, filter: $filter, first: $first) {
+    nodes {
+      id
+      identifier
+      title
+      description
+    }
+  }
+}
+"""
+
+CUSTOMER_UPSERT_MUTATION = """
+mutation CustomerUpsert($input: CustomerUpsertInput!) {
+  customerUpsert(input: $input) {
+    success
+    customer {
+      id
+      name
+      externalIds
+    }
+  }
+}
+"""
+
+ISSUE_CREATE_MUTATION = """
+mutation IssueCreate($input: IssueCreateInput!) {
+  issueCreate(input: $input) {
+    success
+    issue {
+      id
+      identifier
+      title
+      url
+    }
+  }
+}
+"""
+
+CUSTOMER_NEED_CREATE_MUTATION = """
+mutation CustomerNeedCreate($input: CustomerNeedCreateInput!) {
+  customerNeedCreate(input: $input) {
+    success
+    need {
+      id
+      body
+      customer {
+        id
+        name
+      }
+      issue {
+        id
+        identifier
+        title
+        url
+      }
+    }
+  }
+}
+"""
+
+
+_settings: Settings | None = None
+
+
+def _get_settings() -> Settings:
+    global _settings
+    if _settings is None:
+        _settings = Settings()
+    return _settings
+
+
+def _get_linear_config() -> tuple[LinearClient, str, str]:
+    """Return a configured Linear client, project ID, and team ID.
+
+    Raises RuntimeError if any required setting is missing.
+    """
+    secrets = _get_settings().secrets
+    if not secrets.linear_api_key:
+        raise RuntimeError("LINEAR_API_KEY is not configured")
+    if not secrets.linear_feature_request_project_id:
+        raise RuntimeError("LINEAR_FEATURE_REQUEST_PROJECT_ID is not configured")
+    if not secrets.linear_feature_request_team_id:
+        raise RuntimeError("LINEAR_FEATURE_REQUEST_TEAM_ID is not configured")
+
+    credentials = APIKeyCredentials(
+        id="system-linear",
+        provider="linear",
+        api_key=SecretStr(secrets.linear_api_key),
+        title="System Linear API Key",
+    )
+    client = LinearClient(credentials=credentials)
+    return (
+        client,
+        secrets.linear_feature_request_project_id,
+        secrets.linear_feature_request_team_id,
+    )
+
+
+class SearchFeatureRequestsTool(BaseTool):
+    """Tool for searching existing feature requests in Linear."""
+
+    @property
+    def name(self) -> str:
+        return "search_feature_requests"
+
+    @property
+    def description(self) -> str:
+        return (
+            "Search existing feature requests to check if a similar request "
+            "already exists before creating a new one. Returns matching feature "
+            "requests with their ID, title, and description."
+        )
+
+    @property
+    def parameters(self) -> dict[str, Any]:
+        return {
+            "type": "object",
+            "properties": {
+                "query": {
+                    "type": "string",
+                    "description": "Search term to find matching feature requests.",
+                },
+            },
+            "required": ["query"],
+        }
+
+    @property
+    def requires_auth(self) -> bool:
+        return True
+
+    async def _execute(
+        self,
+        user_id: str | None,
+        session: ChatSession,
+        **kwargs,
+    ) -> ToolResponseBase:
+        query = kwargs.get("query", "").strip()
+        session_id = session.session_id if session else None
+
+        if not query:
+            return ErrorResponse(
+                message="Please provide a search query.",
+                error="Missing query parameter",
+                session_id=session_id,
+            )
+
+        try:
+            client, project_id, _team_id = _get_linear_config()
+            data = await client.query(
+                SEARCH_ISSUES_QUERY,
+                {
+                    "term": query,
+                    "filter": {
+                        "project": {"id": {"eq": project_id}},
+                    },
+                    "first": MAX_SEARCH_RESULTS,
+                },
+            )
+
+            nodes = data.get("searchIssues", {}).get("nodes", [])
+
+            if not nodes:
+                return NoResultsResponse(
+                    message=f"No feature requests found matching '{query}'.",
+                    suggestions=[
+                        "Try different keywords",
+                        "Use broader search terms",
+                        "You can create a new feature request if none exists",
+                    ],
+                    session_id=session_id,
+                )
+
+            results = [
+                FeatureRequestInfo(
+                    id=node["id"],
+                    identifier=node["identifier"],
+                    title=node["title"],
+                    description=node.get("description"),
+                )
+                for node in nodes
+            ]
+
+            return FeatureRequestSearchResponse(
+                message=f"Found {len(results)} feature request(s) matching '{query}'.",
+                results=results,
+                count=len(results),
+                query=query,
+                session_id=session_id,
+            )
+        except Exception as e:
+            logger.exception("Failed to search feature requests")
+            return ErrorResponse(
+                message="Failed to search feature requests.",
+                error=str(e),
+                session_id=session_id,
+            )
+
+
+class CreateFeatureRequestTool(BaseTool):
+    """Tool for creating feature requests (or adding needs to existing ones)."""
+
+    @property
+    def name(self) -> str:
+        return "create_feature_request"
+
+    @property
+    def description(self) -> str:
+        return (
+            "Create a new feature request or add a customer need to an existing one. "
+            "Always search first with search_feature_requests to avoid duplicates. "
+            "If a matching request exists, pass its ID as existing_issue_id to add "
+            "the user's need to it instead of creating a duplicate."
+        )
+
+    @property
+    def parameters(self) -> dict[str, Any]:
+        return {
+            "type": "object",
+            "properties": {
+                "title": {
+                    "type": "string",
+                    "description": "Title for the feature request.",
+                },
+                "description": {
+                    "type": "string",
+                    "description": "Detailed description of what the user wants and why.",
+                },
+                "existing_issue_id": {
+                    "type": "string",
+                    "description": (
+                        "If adding a need to an existing feature request, "
+                        "provide its Linear issue ID (from search results). "
+                        "Omit to create a new feature request."
+                    ),
+                },
+            },
+            "required": ["title", "description"],
+        }
+
+    @property
+    def requires_auth(self) -> bool:
+        return True
+
+    async def _find_or_create_customer(
+        self, client: LinearClient, user_id: str, name: str
+    ) -> dict:
+        """Find existing customer by user_id or create a new one via upsert.
+
+        Args:
+            client: Linear API client.
+            user_id: Stable external ID used to deduplicate customers.
+            name: Human-readable display name (e.g. the user's email).
+        """
+        data = await client.mutate(
+            CUSTOMER_UPSERT_MUTATION,
+            {
+                "input": {
+                    "name": name,
+                    "externalId": user_id,
+                },
+            },
+        )
+        result = data.get("customerUpsert", {})
+        if not result.get("success"):
+            raise RuntimeError(f"Failed to upsert customer: {data}")
+        return result["customer"]
+
+    async def _execute(
+        self,
+        user_id: str | None,
+        session: ChatSession,
+        **kwargs,
+    ) -> ToolResponseBase:
+        title = kwargs.get("title", "").strip()
+        description = kwargs.get("description", "").strip()
+        existing_issue_id = kwargs.get("existing_issue_id")
+        session_id = session.session_id if session else None
+
+        if not title or not description:
+            return ErrorResponse(
+                message="Both title and description are required.",
+                error="Missing required parameters",
+                session_id=session_id,
+            )
+
+        if not user_id:
+            return ErrorResponse(
+                message="Authentication required to create feature requests.",
+                error="Missing user_id",
+                session_id=session_id,
+            )
+
+        try:
+            client, project_id, team_id = _get_linear_config()
+        except Exception as e:
+            logger.exception("Failed to initialize Linear client")
+            return ErrorResponse(
+                message="Failed to create feature request.",
+                error=str(e),
+                session_id=session_id,
+            )
+
+        # Resolve a human-readable name (email) for the Linear customer record.
+        # Fall back to user_id if the lookup fails or returns None.
+        try:
+            customer_display_name = await get_user_email_by_id(user_id) or user_id
+        except Exception:
+            customer_display_name = user_id
+
+        # Step 1: Find or create customer for this user
+        try:
+            customer = await self._find_or_create_customer(
+                client, user_id, customer_display_name
+            )
+            customer_id = customer["id"]
+            customer_name = customer["name"]
+        except Exception as e:
+            logger.exception("Failed to upsert customer in Linear")
+            return ErrorResponse(
+                message="Failed to create feature request.",
+                error=str(e),
+                session_id=session_id,
+            )
+
+        # Step 2: Create or reuse issue
+        issue_id: str | None = None
+        issue_identifier: str | None = None
+        if existing_issue_id:
+            # Add need to existing issue - we still need the issue details for response
+            is_new_issue = False
+            issue_id = existing_issue_id
+        else:
+            # Create new issue in the feature requests project
+            try:
+                data = await client.mutate(
+                    ISSUE_CREATE_MUTATION,
+                    {
+                        "input": {
+                            "title": title,
+                            "description": description,
+                            "teamId": team_id,
+                            "projectId": project_id,
+                        },
+                    },
+                )
+                result = data.get("issueCreate", {})
+                if not result.get("success"):
+                    return ErrorResponse(
+                        message="Failed to create feature request issue.",
+                        error=str(data),
+                        session_id=session_id,
+                    )
+                issue = result["issue"]
+                issue_id = issue["id"]
+                issue_identifier = issue.get("identifier")
+            except Exception as e:
+                logger.exception("Failed to create feature request issue")
+                return ErrorResponse(
+                    message="Failed to create feature request.",
+                    error=str(e),
+                    session_id=session_id,
+                )
+            is_new_issue = True
+
+        # Step 3: Create customer need on the issue
+        try:
+            data = await client.mutate(
+                CUSTOMER_NEED_CREATE_MUTATION,
+                {
+                    "input": {
+                        "customerId": customer_id,
+                        "issueId": issue_id,
+                        "body": description,
+                        "priority": 0,
+                    },
+                },
+            )
+            need_result = data.get("customerNeedCreate", {})
+            if not need_result.get("success"):
+                orphaned = (
+                    {"issue_id": issue_id, "issue_identifier": issue_identifier}
+                    if is_new_issue
+                    else None
+                )
+                return ErrorResponse(
+                    message="Failed to attach customer need to the feature request.",
+                    error=str(data),
+                    details=orphaned,
+                    session_id=session_id,
+                )
+            need = need_result["need"]
+            issue_info = need["issue"]
+        except Exception as e:
+            logger.exception("Failed to create customer need")
+            orphaned = (
+                {"issue_id": issue_id, "issue_identifier": issue_identifier}
+                if is_new_issue
+                else None
+            )
+            return ErrorResponse(
+                message="Failed to attach customer need to the feature request.",
+                error=str(e),
+                details=orphaned,
+                session_id=session_id,
+            )
+
+        return FeatureRequestCreatedResponse(
+            message=(
+                f"{'Created new feature request' if is_new_issue else 'Added your request to existing feature request'}: "
+                f"{issue_info['title']}."
+            ),
+            issue_id=issue_info["id"],
+            issue_identifier=issue_info["identifier"],
+            issue_title=issue_info["title"],
+            issue_url=issue_info.get("url", ""),
+            is_new_issue=is_new_issue,
+            customer_name=customer_name,
+            session_id=session_id,
+        )

autogpt_platform/backend/backend/api/features/chat/tools/feature_requests_test.py

@@ -0,0 +1,615 @@
+"""Tests for SearchFeatureRequestsTool and CreateFeatureRequestTool."""
+
+from unittest.mock import AsyncMock, patch
+
+import pytest
+
+from backend.api.features.chat.tools.feature_requests import (
+    CreateFeatureRequestTool,
+    SearchFeatureRequestsTool,
+)
+from backend.api.features.chat.tools.models import (
+    ErrorResponse,
+    FeatureRequestCreatedResponse,
+    FeatureRequestSearchResponse,
+    NoResultsResponse,
+)
+
+from ._test_data import make_session
+
+_TEST_USER_ID = "test-user-feature-requests"
+_TEST_USER_EMAIL = "testuser@example.com"
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+
+_FAKE_PROJECT_ID = "test-project-id"
+_FAKE_TEAM_ID = "test-team-id"
+
+
+def _mock_linear_config(*, query_return=None, mutate_return=None):
+    """Return a patched _get_linear_config that yields a mock LinearClient."""
+    client = AsyncMock()
+    if query_return is not None:
+        client.query.return_value = query_return
+    if mutate_return is not None:
+        client.mutate.return_value = mutate_return
+    return (
+        patch(
+            "backend.api.features.chat.tools.feature_requests._get_linear_config",
+            return_value=(client, _FAKE_PROJECT_ID, _FAKE_TEAM_ID),
+        ),
+        client,
+    )
+
+
+def _search_response(nodes: list[dict]) -> dict:
+    return {"searchIssues": {"nodes": nodes}}
+
+
+def _customer_upsert_response(
+    customer_id: str = "cust-1", name: str = _TEST_USER_EMAIL, success: bool = True
+) -> dict:
+    return {
+        "customerUpsert": {
+            "success": success,
+            "customer": {"id": customer_id, "name": name, "externalIds": [name]},
+        }
+    }
+
+
+def _issue_create_response(
+    issue_id: str = "issue-1",
+    identifier: str = "FR-1",
+    title: str = "New Feature",
+    success: bool = True,
+) -> dict:
+    return {
+        "issueCreate": {
+            "success": success,
+            "issue": {
+                "id": issue_id,
+                "identifier": identifier,
+                "title": title,
+                "url": f"https://linear.app/issue/{identifier}",
+            },
+        }
+    }
+
+
+def _need_create_response(
+    need_id: str = "need-1",
+    issue_id: str = "issue-1",
+    identifier: str = "FR-1",
+    title: str = "New Feature",
+    success: bool = True,
+) -> dict:
+    return {
+        "customerNeedCreate": {
+            "success": success,
+            "need": {
+                "id": need_id,
+                "body": "description",
+                "customer": {"id": "cust-1", "name": _TEST_USER_EMAIL},
+                "issue": {
+                    "id": issue_id,
+                    "identifier": identifier,
+                    "title": title,
+                    "url": f"https://linear.app/issue/{identifier}",
+                },
+            },
+        }
+    }
+
+
+# ===========================================================================
+# SearchFeatureRequestsTool
+# ===========================================================================
+
+
+class TestSearchFeatureRequestsTool:
+    """Tests for SearchFeatureRequestsTool._execute."""
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_successful_search(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        nodes = [
+            {
+                "id": "id-1",
+                "identifier": "FR-1",
+                "title": "Dark mode",
+                "description": "Add dark mode support",
+            },
+            {
+                "id": "id-2",
+                "identifier": "FR-2",
+                "title": "Dark theme",
+                "description": None,
+            },
+        ]
+        patcher, _ = _mock_linear_config(query_return=_search_response(nodes))
+        with patcher:
+            tool = SearchFeatureRequestsTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID, session=session, query="dark mode"
+            )
+
+        assert isinstance(resp, FeatureRequestSearchResponse)
+        assert resp.count == 2
+        assert resp.results[0].id == "id-1"
+        assert resp.results[1].identifier == "FR-2"
+        assert resp.query == "dark mode"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_no_results(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, _ = _mock_linear_config(query_return=_search_response([]))
+        with patcher:
+            tool = SearchFeatureRequestsTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID, session=session, query="nonexistent"
+            )
+
+        assert isinstance(resp, NoResultsResponse)
+        assert "nonexistent" in resp.message
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_empty_query_returns_error(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        tool = SearchFeatureRequestsTool()
+        resp = await tool._execute(user_id=_TEST_USER_ID, session=session, query="   ")
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.error is not None
+        assert "query" in resp.error.lower()
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_missing_query_returns_error(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        tool = SearchFeatureRequestsTool()
+        resp = await tool._execute(user_id=_TEST_USER_ID, session=session)
+
+        assert isinstance(resp, ErrorResponse)
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_api_failure(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        client.query.side_effect = RuntimeError("Linear API down")
+        with patcher:
+            tool = SearchFeatureRequestsTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID, session=session, query="test"
+            )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.error is not None
+        assert "Linear API down" in resp.error
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_malformed_node_returns_error(self):
+        """A node missing required keys should be caught by the try/except."""
+        session = make_session(user_id=_TEST_USER_ID)
+        # Node missing 'identifier' key
+        bad_nodes = [{"id": "id-1", "title": "Missing identifier"}]
+        patcher, _ = _mock_linear_config(query_return=_search_response(bad_nodes))
+        with patcher:
+            tool = SearchFeatureRequestsTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID, session=session, query="test"
+            )
+
+        assert isinstance(resp, ErrorResponse)
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_linear_client_init_failure(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        with patch(
+            "backend.api.features.chat.tools.feature_requests._get_linear_config",
+            side_effect=RuntimeError("No API key"),
+        ):
+            tool = SearchFeatureRequestsTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID, session=session, query="test"
+            )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.error is not None
+        assert "No API key" in resp.error
+
+
+# ===========================================================================
+# CreateFeatureRequestTool
+# ===========================================================================
+
+
+class TestCreateFeatureRequestTool:
+    """Tests for CreateFeatureRequestTool._execute."""
+
+    @pytest.fixture(autouse=True)
+    def _patch_email_lookup(self):
+        with patch(
+            "backend.api.features.chat.tools.feature_requests.get_user_email_by_id",
+            new_callable=AsyncMock,
+            return_value=_TEST_USER_EMAIL,
+        ):
+            yield
+
+    # ---- Happy paths -------------------------------------------------------
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_create_new_issue(self):
+        """Full happy path: upsert customer -> create issue -> attach need."""
+        session = make_session(user_id=_TEST_USER_ID)
+
+        patcher, client = _mock_linear_config()
+        client.mutate.side_effect = [
+            _customer_upsert_response(),
+            _issue_create_response(),
+            _need_create_response(),
+        ]
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="New Feature",
+                description="Please add this",
+            )
+
+        assert isinstance(resp, FeatureRequestCreatedResponse)
+        assert resp.is_new_issue is True
+        assert resp.issue_identifier == "FR-1"
+        assert resp.customer_name == _TEST_USER_EMAIL
+        assert client.mutate.call_count == 3
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_add_need_to_existing_issue(self):
+        """When existing_issue_id is provided, skip issue creation."""
+        session = make_session(user_id=_TEST_USER_ID)
+
+        patcher, client = _mock_linear_config()
+        client.mutate.side_effect = [
+            _customer_upsert_response(),
+            _need_create_response(issue_id="existing-1", identifier="FR-99"),
+        ]
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Existing Feature",
+                description="Me too",
+                existing_issue_id="existing-1",
+            )
+
+        assert isinstance(resp, FeatureRequestCreatedResponse)
+        assert resp.is_new_issue is False
+        assert resp.issue_id == "existing-1"
+        # Only 2 mutations: customer upsert + need create (no issue create)
+        assert client.mutate.call_count == 2
+
+    # ---- Validation errors -------------------------------------------------
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_missing_title(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        tool = CreateFeatureRequestTool()
+        resp = await tool._execute(
+            user_id=_TEST_USER_ID,
+            session=session,
+            title="",
+            description="some desc",
+        )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.error is not None
+        assert "required" in resp.error.lower()
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_missing_description(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        tool = CreateFeatureRequestTool()
+        resp = await tool._execute(
+            user_id=_TEST_USER_ID,
+            session=session,
+            title="Some title",
+            description="",
+        )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.error is not None
+        assert "required" in resp.error.lower()
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_missing_user_id(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        tool = CreateFeatureRequestTool()
+        resp = await tool._execute(
+            user_id=None,
+            session=session,
+            title="Some title",
+            description="Some desc",
+        )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.error is not None
+        assert "user_id" in resp.error.lower()
+
+    # ---- Linear client init failure ----------------------------------------
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_linear_client_init_failure(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        with patch(
+            "backend.api.features.chat.tools.feature_requests._get_linear_config",
+            side_effect=RuntimeError("No API key"),
+        ):
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.error is not None
+        assert "No API key" in resp.error
+
+    # ---- Customer upsert failures ------------------------------------------
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_customer_upsert_api_error(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        client.mutate.side_effect = RuntimeError("Customer API error")
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.error is not None
+        assert "Customer API error" in resp.error
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_customer_upsert_not_success(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        client.mutate.return_value = _customer_upsert_response(success=False)
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_customer_malformed_response(self):
+        """Customer dict missing 'id' key should be caught."""
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        # success=True but customer has no 'id'
+        client.mutate.return_value = {
+            "customerUpsert": {
+                "success": True,
+                "customer": {"name": _TEST_USER_ID},
+            }
+        }
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+
+    # ---- Issue creation failures -------------------------------------------
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_issue_create_api_error(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        client.mutate.side_effect = [
+            _customer_upsert_response(),
+            RuntimeError("Issue create failed"),
+        ]
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.error is not None
+        assert "Issue create failed" in resp.error
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_issue_create_not_success(self):
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        client.mutate.side_effect = [
+            _customer_upsert_response(),
+            _issue_create_response(success=False),
+        ]
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+        assert "Failed to create feature request issue" in resp.message
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_issue_create_malformed_response(self):
+        """issueCreate success=True but missing 'issue' key."""
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        client.mutate.side_effect = [
+            _customer_upsert_response(),
+            {"issueCreate": {"success": True}},  # no 'issue' key
+        ]
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+
+    # ---- Customer need attachment failures ---------------------------------
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_need_create_api_error_new_issue(self):
+        """Need creation fails after new issue was created -> orphaned issue info."""
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        client.mutate.side_effect = [
+            _customer_upsert_response(),
+            _issue_create_response(issue_id="orphan-1", identifier="FR-10"),
+            RuntimeError("Need attach failed"),
+        ]
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.error is not None
+        assert "Need attach failed" in resp.error
+        assert resp.details is not None
+        assert resp.details["issue_id"] == "orphan-1"
+        assert resp.details["issue_identifier"] == "FR-10"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_need_create_api_error_existing_issue(self):
+        """Need creation fails on existing issue -> no orphaned info."""
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        client.mutate.side_effect = [
+            _customer_upsert_response(),
+            RuntimeError("Need attach failed"),
+        ]
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+                existing_issue_id="existing-1",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.details is None
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_need_create_not_success_includes_orphaned_info(self):
+        """customerNeedCreate returns success=False -> includes orphaned issue."""
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        client.mutate.side_effect = [
+            _customer_upsert_response(),
+            _issue_create_response(issue_id="orphan-2", identifier="FR-20"),
+            _need_create_response(success=False),
+        ]
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.details is not None
+        assert resp.details["issue_id"] == "orphan-2"
+        assert resp.details["issue_identifier"] == "FR-20"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_need_create_not_success_existing_issue_no_details(self):
+        """customerNeedCreate fails on existing issue -> no orphaned info."""
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        client.mutate.side_effect = [
+            _customer_upsert_response(),
+            _need_create_response(success=False),
+        ]
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+                existing_issue_id="existing-1",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.details is None
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_need_create_malformed_response(self):
+        """need_result missing 'need' key after success=True."""
+        session = make_session(user_id=_TEST_USER_ID)
+        patcher, client = _mock_linear_config()
+        client.mutate.side_effect = [
+            _customer_upsert_response(),
+            _issue_create_response(),
+            {"customerNeedCreate": {"success": True}},  # no 'need' key
+        ]
+
+        with patcher:
+            tool = CreateFeatureRequestTool()
+            resp = await tool._execute(
+                user_id=_TEST_USER_ID,
+                session=session,
+                title="Title",
+                description="Desc",
+            )
+
+        assert isinstance(resp, ErrorResponse)
+        assert resp.details is not None
+        assert resp.details["issue_id"] == "issue-1"

autogpt_platform/backend/backend/api/features/chat/tools/find_block.py

@@ -146,6 +146,7 @@ class FindBlockTool(BaseTool):
                         id=block_id,
                         name=block.name,
                         description=block.description or "",
+                        categories=[c.value for c in block.categories],
                     )
                 )
 

autogpt_platform/backend/backend/api/features/chat/tools/models.py

@@ -41,6 +41,15 @@ class ResponseType(str, Enum):
     OPERATION_IN_PROGRESS = "operation_in_progress"
     # Input validation
     INPUT_VALIDATION_ERROR = "input_validation_error"
+    # Web fetch
+    WEB_FETCH = "web_fetch"
+    # Code execution
+    BASH_EXEC = "bash_exec"
+    # Operation status check
+    OPERATION_STATUS = "operation_status"
+    # Feature request types
+    FEATURE_REQUEST_SEARCH = "feature_request_search"
+    FEATURE_REQUEST_CREATED = "feature_request_created"
 
 
 # Base response model
@@ -335,6 +344,19 @@ class BlockInfoSummary(BaseModel):
     id: str
     name: str
     description: str
+    categories: list[str]
+    input_schema: dict[str, Any] = Field(
+        default_factory=dict,
+        description="Full JSON schema for block inputs",
+    )
+    output_schema: dict[str, Any] = Field(
+        default_factory=dict,
+        description="Full JSON schema for block outputs",
+    )
+    required_inputs: list[BlockInputFieldInfo] = Field(
+        default_factory=list,
+        description="List of input fields for this block",
+    )
 
 
 class BlockListResponse(ToolResponseBase):
@@ -344,6 +366,10 @@ class BlockListResponse(ToolResponseBase):
     blocks: list[BlockInfoSummary]
     count: int
     query: str
+    usage_hint: str = Field(
+        default="To execute a block, call run_block with block_id set to the block's "
+        "'id' field and input_data containing the fields listed in required_inputs."
+    )
 
 
 class BlockDetails(BaseModel):
@@ -430,3 +456,55 @@ class AsyncProcessingResponse(ToolResponseBase):
     status: str = "accepted"  # Must be "accepted" for detection
     operation_id: str | None = None
     task_id: str | None = None
+
+
+class WebFetchResponse(ToolResponseBase):
+    """Response for web_fetch tool."""
+
+    type: ResponseType = ResponseType.WEB_FETCH
+    url: str
+    status_code: int
+    content_type: str
+    content: str
+    truncated: bool = False
+
+
+class BashExecResponse(ToolResponseBase):
+    """Response for bash_exec tool."""
+
+    type: ResponseType = ResponseType.BASH_EXEC
+    stdout: str
+    stderr: str
+    exit_code: int
+    timed_out: bool = False
+
+
+# Feature request models
+class FeatureRequestInfo(BaseModel):
+    """Information about a feature request issue."""
+
+    id: str
+    identifier: str
+    title: str
+    description: str | None = None
+
+
+class FeatureRequestSearchResponse(ToolResponseBase):
+    """Response for search_feature_requests tool."""
+
+    type: ResponseType = ResponseType.FEATURE_REQUEST_SEARCH
+    results: list[FeatureRequestInfo]
+    count: int
+    query: str
+
+
+class FeatureRequestCreatedResponse(ToolResponseBase):
+    """Response for create_feature_request tool."""
+
+    type: ResponseType = ResponseType.FEATURE_REQUEST_CREATED
+    issue_id: str
+    issue_identifier: str
+    issue_title: str
+    issue_url: str
+    is_new_issue: bool  # False if added to existing
+    customer_name: str

autogpt_platform/backend/backend/api/features/chat/tools/sandbox.py

@@ -0,0 +1,265 @@
+"""Sandbox execution utilities for code execution tools.
+
+Provides filesystem + network isolated command execution using **bubblewrap**
+(``bwrap``): whitelist-only filesystem (only system dirs visible read-only),
+writable workspace only, clean environment, network blocked.
+
+Tools that call :func:`run_sandboxed` must first check :func:`has_full_sandbox`
+and refuse to run if bubblewrap is not available.
+"""
+
+import asyncio
+import logging
+import os
+import platform
+import shutil
+
+logger = logging.getLogger(__name__)
+
+_DEFAULT_TIMEOUT = 30
+_MAX_TIMEOUT = 120
+
+
+# ---------------------------------------------------------------------------
+# Sandbox capability detection (cached at first call)
+# ---------------------------------------------------------------------------
+
+_BWRAP_AVAILABLE: bool | None = None
+
+
+def has_full_sandbox() -> bool:
+    """Return True if bubblewrap is available (filesystem + network isolation).
+
+    On non-Linux platforms (macOS), always returns False.
+    """
+    global _BWRAP_AVAILABLE
+    if _BWRAP_AVAILABLE is None:
+        _BWRAP_AVAILABLE = (
+            platform.system() == "Linux" and shutil.which("bwrap") is not None
+        )
+    return _BWRAP_AVAILABLE
+
+
+WORKSPACE_PREFIX = "/tmp/copilot-"
+
+
+def make_session_path(session_id: str) -> str:
+    """Build a sanitized, session-specific path under :data:`WORKSPACE_PREFIX`.
+
+    Shared by both the SDK working-directory setup and the sandbox tools so
+    they always resolve to the same directory for a given session.
+
+    Steps:
+        1. Strip all characters except ``[A-Za-z0-9-]``.
+        2. Construct ``/tmp/copilot-<safe_id>``.
+        3. Validate via ``os.path.normpath`` + ``startswith`` (CodeQL-recognised
+           sanitizer) to prevent path traversal.
+
+    Raises:
+        ValueError: If the resulting path escapes the prefix.
+    """
+    import re
+
+    safe_id = re.sub(r"[^A-Za-z0-9-]", "", session_id)
+    if not safe_id:
+        safe_id = "default"
+    path = os.path.normpath(f"{WORKSPACE_PREFIX}{safe_id}")
+    if not path.startswith(WORKSPACE_PREFIX):
+        raise ValueError(f"Session path escaped prefix: {path}")
+    return path
+
+
+def get_workspace_dir(session_id: str) -> str:
+    """Get or create the workspace directory for a session.
+
+    Uses :func:`make_session_path` — the same path the SDK uses — so that
+    bash_exec shares the workspace with the SDK file tools.
+    """
+    workspace = make_session_path(session_id)
+    os.makedirs(workspace, exist_ok=True)
+    return workspace
+
+
+# ---------------------------------------------------------------------------
+# Bubblewrap command builder
+# ---------------------------------------------------------------------------
+
+# System directories mounted read-only inside the sandbox.
+# ONLY these are visible — /app, /root, /home, /opt, /var etc. are NOT accessible.
+_SYSTEM_RO_BINDS = [
+    "/usr",  # binaries, libraries, Python interpreter
+    "/etc",  # system config: ld.so, locale, passwd, alternatives
+]
+
+# Compat paths: symlinks to /usr/* on modern Debian, real dirs on older systems.
+# On Debian 13 these are symlinks (e.g. /bin -> usr/bin).  bwrap --ro-bind
+# can't create a symlink target, so we detect and use --symlink instead.
+# /lib64 is critical: the ELF dynamic linker lives at /lib64/ld-linux-x86-64.so.2.
+_COMPAT_PATHS = [
+    ("/bin", "usr/bin"),  # -> /usr/bin on Debian 13
+    ("/sbin", "usr/sbin"),  # -> /usr/sbin on Debian 13
+    ("/lib", "usr/lib"),  # -> /usr/lib on Debian 13
+    ("/lib64", "usr/lib64"),  # 64-bit libraries / ELF interpreter
+]
+
+# Resource limits to prevent fork bombs, memory exhaustion, and disk abuse.
+# Applied via ulimit inside the sandbox before exec'ing the user command.
+_RESOURCE_LIMITS = (
+    "ulimit -u 64"  # max 64 processes  (prevents fork bombs)
+    " -v 524288"  # 512 MB virtual memory
+    " -f 51200"  # 50 MB max file size  (1024-byte blocks)
+    " -n 256"  # 256 open file descriptors
+    " 2>/dev/null"
+)
+
+
+def _build_bwrap_command(
+    command: list[str], cwd: str, env: dict[str, str]
+) -> list[str]:
+    """Build a bubblewrap command with strict filesystem + network isolation.
+
+    Security model:
+    - **Whitelist-only filesystem**: only system directories (``/usr``, ``/etc``,
+      ``/bin``, ``/lib``) are mounted read-only.  Application code (``/app``),
+      home directories, ``/var``, ``/opt``, etc. are NOT accessible at all.
+    - **Writable workspace only**: the per-session workspace is the sole
+      writable path.
+    - **Clean environment**: ``--clearenv`` wipes all inherited env vars.
+      Only the explicitly-passed safe env vars are set inside the sandbox.
+    - **Network isolation**: ``--unshare-net`` blocks all network access.
+    - **Resource limits**: ulimit caps on processes (64), memory (512MB),
+      file size (50MB), and open FDs (256) to prevent fork bombs and abuse.
+    - **New session**: prevents terminal control escape.
+    - **Die with parent**: prevents orphaned sandbox processes.
+    """
+    cmd = [
+        "bwrap",
+        # Create a new user namespace so bwrap can set up sandboxing
+        # inside unprivileged Docker containers (no CAP_SYS_ADMIN needed).
+        "--unshare-user",
+        # Wipe all inherited environment variables (API keys, secrets, etc.)
+        "--clearenv",
+    ]
+
+    # Set only the safe env vars inside the sandbox
+    for key, value in env.items():
+        cmd.extend(["--setenv", key, value])
+
+    # System directories: read-only
+    for path in _SYSTEM_RO_BINDS:
+        cmd.extend(["--ro-bind", path, path])
+
+    # Compat paths: use --symlink when host path is a symlink (Debian 13),
+    # --ro-bind when it's a real directory (older distros).
+    for path, symlink_target in _COMPAT_PATHS:
+        if os.path.islink(path):
+            cmd.extend(["--symlink", symlink_target, path])
+        elif os.path.exists(path):
+            cmd.extend(["--ro-bind", path, path])
+
+    # Wrap the user command with resource limits:
+    #   sh -c 'ulimit ...; exec "$@"' -- <original command>
+    # `exec "$@"` replaces the shell so there's no extra process overhead,
+    # and properly handles arguments with spaces.
+    limited_command = [
+        "sh",
+        "-c",
+        f'{_RESOURCE_LIMITS}; exec "$@"',
+        "--",
+        *command,
+    ]
+
+    cmd.extend(
+        [
+            # Fresh virtual filesystems
+            "--dev",
+            "/dev",
+            "--proc",
+            "/proc",
+            "--tmpfs",
+            "/tmp",
+            # Workspace bind AFTER --tmpfs /tmp so it's visible through the tmpfs.
+            # (workspace lives under /tmp/copilot-<session>)
+            "--bind",
+            cwd,
+            cwd,
+            # Isolation
+            "--unshare-net",
+            "--die-with-parent",
+            "--new-session",
+            "--chdir",
+            cwd,
+            "--",
+            *limited_command,
+        ]
+    )
+
+    return cmd
+
+
+# ---------------------------------------------------------------------------
+# Public API
+# ---------------------------------------------------------------------------
+
+
+async def run_sandboxed(
+    command: list[str],
+    cwd: str,
+    timeout: int = _DEFAULT_TIMEOUT,
+    env: dict[str, str] | None = None,
+) -> tuple[str, str, int, bool]:
+    """Run a command inside a bubblewrap sandbox.
+
+    Callers **must** check :func:`has_full_sandbox` before calling this
+    function.  If bubblewrap is not available, this function raises
+    :class:`RuntimeError` rather than running unsandboxed.
+
+    Returns:
+        (stdout, stderr, exit_code, timed_out)
+    """
+    if not has_full_sandbox():
+        raise RuntimeError(
+            "run_sandboxed() requires bubblewrap but bwrap is not available. "
+            "Callers must check has_full_sandbox() before calling this function."
+        )
+
+    timeout = min(max(timeout, 1), _MAX_TIMEOUT)
+
+    safe_env = {
+        "PATH": "/usr/local/bin:/usr/bin:/bin",
+        "HOME": cwd,
+        "TMPDIR": cwd,
+        "LANG": "en_US.UTF-8",
+        "PYTHONDONTWRITEBYTECODE": "1",
+        "PYTHONIOENCODING": "utf-8",
+    }
+    if env:
+        safe_env.update(env)
+
+    full_command = _build_bwrap_command(command, cwd, safe_env)
+
+    try:
+        proc = await asyncio.create_subprocess_exec(
+            *full_command,
+            stdout=asyncio.subprocess.PIPE,
+            stderr=asyncio.subprocess.PIPE,
+            cwd=cwd,
+            env=safe_env,
+        )
+
+        try:
+            stdout_bytes, stderr_bytes = await asyncio.wait_for(
+                proc.communicate(), timeout=timeout
+            )
+            stdout = stdout_bytes.decode("utf-8", errors="replace")
+            stderr = stderr_bytes.decode("utf-8", errors="replace")
+            return stdout, stderr, proc.returncode or 0, False
+        except asyncio.TimeoutError:
+            proc.kill()
+            await proc.communicate()
+            return "", f"Execution timed out after {timeout}s", -1, True
+
+    except RuntimeError:
+        raise
+    except Exception as e:
+        return "", f"Sandbox error: {e}", -1, False

autogpt_platform/backend/backend/api/features/chat/tools/utils.py

@@ -15,6 +15,7 @@ from backend.data.model import (
     OAuth2Credentials,
 )
 from backend.integrations.creds_manager import IntegrationCredentialsManager
+from backend.integrations.providers import ProviderName
 from backend.util.exceptions import NotFoundError
 
 logger = logging.getLogger(__name__)
@@ -359,7 +360,7 @@ async def match_user_credentials_to_graph(
         _,
         _,
     ) in aggregated_creds.items():
-        # Find first matching credential by provider, type, and scopes
+        # Find first matching credential by provider, type, scopes, and host/URL
         matching_cred = next(
             (
                 cred
@@ -374,6 +375,10 @@ async def match_user_credentials_to_graph(
                     cred.type != "host_scoped"
                     or _credential_is_for_host(cred, credential_requirements)
                 )
+                and (
+                    cred.provider != ProviderName.MCP
+                    or _credential_is_for_mcp_server(cred, credential_requirements)
+                )
             ),
             None,
         )
@@ -444,6 +449,22 @@ def _credential_is_for_host(
     return credential.matches_url(list(requirements.discriminator_values)[0])
 
 
+def _credential_is_for_mcp_server(
+    credential: Credentials,
+    requirements: CredentialsFieldInfo,
+) -> bool:
+    """Check if an MCP OAuth credential matches the required server URL."""
+    if not requirements.discriminator_values:
+        return True
+
+    server_url = (
+        credential.metadata.get("mcp_server_url")
+        if isinstance(credential, OAuth2Credentials)
+        else None
+    )
+    return server_url in requirements.discriminator_values if server_url else False
+
+
 async def check_user_has_required_credentials(
     user_id: str,
     required_credentials: list[CredentialsMetaInput],

autogpt_platform/backend/backend/api/features/chat/tools/web_fetch.py

@@ -0,0 +1,151 @@
+"""Web fetch tool — safely retrieve public web page content."""
+
+import logging
+from typing import Any
+
+import aiohttp
+import html2text
+
+from backend.api.features.chat.model import ChatSession
+from backend.api.features.chat.tools.base import BaseTool
+from backend.api.features.chat.tools.models import (
+    ErrorResponse,
+    ToolResponseBase,
+    WebFetchResponse,
+)
+from backend.util.request import Requests
+
+logger = logging.getLogger(__name__)
+
+# Limits
+_MAX_CONTENT_BYTES = 102_400  # 100 KB download cap
+_REQUEST_TIMEOUT = aiohttp.ClientTimeout(total=15)
+
+# Content types we'll read as text
+_TEXT_CONTENT_TYPES = {
+    "text/html",
+    "text/plain",
+    "text/xml",
+    "text/csv",
+    "text/markdown",
+    "application/json",
+    "application/xml",
+    "application/xhtml+xml",
+    "application/rss+xml",
+    "application/atom+xml",
+}
+
+
+def _is_text_content(content_type: str) -> bool:
+    base = content_type.split(";")[0].strip().lower()
+    return base in _TEXT_CONTENT_TYPES or base.startswith("text/")
+
+
+def _html_to_text(html: str) -> str:
+    h = html2text.HTML2Text()
+    h.ignore_links = False
+    h.ignore_images = True
+    h.body_width = 0
+    return h.handle(html)
+
+
+class WebFetchTool(BaseTool):
+    """Safely fetch content from a public URL using SSRF-protected HTTP."""
+
+    @property
+    def name(self) -> str:
+        return "web_fetch"
+
+    @property
+    def description(self) -> str:
+        return (
+            "Fetch the content of a public web page by URL. "
+            "Returns readable text extracted from HTML by default. "
+            "Useful for reading documentation, articles, and API responses. "
+            "Only supports HTTP/HTTPS GET requests to public URLs "
+            "(private/internal network addresses are blocked)."
+        )
+
+    @property
+    def parameters(self) -> dict[str, Any]:
+        return {
+            "type": "object",
+            "properties": {
+                "url": {
+                    "type": "string",
+                    "description": "The public HTTP/HTTPS URL to fetch.",
+                },
+                "extract_text": {
+                    "type": "boolean",
+                    "description": (
+                        "If true (default), extract readable text from HTML. "
+                        "If false, return raw content."
+                    ),
+                    "default": True,
+                },
+            },
+            "required": ["url"],
+        }
+
+    @property
+    def requires_auth(self) -> bool:
+        return False
+
+    async def _execute(
+        self,
+        user_id: str | None,
+        session: ChatSession,
+        **kwargs: Any,
+    ) -> ToolResponseBase:
+        url: str = (kwargs.get("url") or "").strip()
+        extract_text: bool = kwargs.get("extract_text", True)
+        session_id = session.session_id if session else None
+
+        if not url:
+            return ErrorResponse(
+                message="Please provide a URL to fetch.",
+                error="missing_url",
+                session_id=session_id,
+            )
+
+        try:
+            client = Requests(raise_for_status=False, retry_max_attempts=1)
+            response = await client.get(url, timeout=_REQUEST_TIMEOUT)
+        except ValueError as e:
+            # validate_url raises ValueError for SSRF / blocked IPs
+            return ErrorResponse(
+                message=f"URL blocked: {e}",
+                error="url_blocked",
+                session_id=session_id,
+            )
+        except Exception as e:
+            logger.warning(f"[web_fetch] Request failed for {url}: {e}")
+            return ErrorResponse(
+                message=f"Failed to fetch URL: {e}",
+                error="fetch_failed",
+                session_id=session_id,
+            )
+
+        content_type = response.headers.get("content-type", "")
+        if not _is_text_content(content_type):
+            return ErrorResponse(
+                message=f"Non-text content type: {content_type.split(';')[0]}",
+                error="unsupported_content_type",
+                session_id=session_id,
+            )
+
+        raw = response.content[:_MAX_CONTENT_BYTES]
+        text = raw.decode("utf-8", errors="replace")
+
+        if extract_text and "html" in content_type.lower():
+            text = _html_to_text(text)
+
+        return WebFetchResponse(
+            message=f"Fetched {url}",
+            url=response.url,
+            status_code=response.status,
+            content_type=content_type.split(";")[0].strip(),
+            content=text,
+            truncated=False,
+            session_id=session_id,
+        )

autogpt_platform/backend/backend/api/features/chat/tools/workspace_files.py

@@ -88,7 +88,9 @@ class ListWorkspaceFilesTool(BaseTool):
     @property
     def description(self) -> str:
         return (
-            "List files in the user's workspace. "
+            "List files in the user's persistent workspace (cloud storage). "
+            "These files survive across sessions. "
+            "For ephemeral session files, use the SDK Read/Glob tools instead. "
             "Returns file names, paths, sizes, and metadata. "
             "Optionally filter by path prefix."
         )
@@ -204,7 +206,9 @@ class ReadWorkspaceFileTool(BaseTool):
     @property
     def description(self) -> str:
         return (
-            "Read a file from the user's workspace. "
+            "Read a file from the user's persistent workspace (cloud storage). "
+            "These files survive across sessions. "
+            "For ephemeral session files, use the SDK Read tool instead. "
             "Specify either file_id or path to identify the file. "
             "For small text files, returns content directly. "
             "For large or binary files, returns metadata and a download URL. "
@@ -378,7 +382,9 @@ class WriteWorkspaceFileTool(BaseTool):
     @property
     def description(self) -> str:
         return (
-            "Write or create a file in the user's workspace. "
+            "Write or create a file in the user's persistent workspace (cloud storage). "
+            "These files survive across sessions. "
+            "For ephemeral session files, use the SDK Write tool instead. "
             "Provide the content as a base64-encoded string. "
             f"Maximum file size is {Config().max_file_size_mb}MB. "
             "Files are saved to the current session's folder by default. "
@@ -523,7 +529,7 @@ class DeleteWorkspaceFileTool(BaseTool):
     @property
     def description(self) -> str:
         return (
-            "Delete a file from the user's workspace. "
+            "Delete a file from the user's persistent workspace (cloud storage). "
             "Specify either file_id or path to identify the file. "
             "Paths are scoped to the current session by default. "
             "Use /sessions/<session_id>/... for cross-session access."

autogpt_platform/backend/backend/api/features/integrations/router.py

@@ -1,7 +1,7 @@
 import asyncio
 import logging
 from datetime import datetime, timedelta, timezone
-from typing import TYPE_CHECKING, Annotated, List, Literal
+from typing import TYPE_CHECKING, Annotated, Any, List, Literal
 
 from autogpt_libs.auth import get_user_id
 from fastapi import (
@@ -14,7 +14,7 @@ from fastapi import (
     Security,
     status,
 )
-from pydantic import BaseModel, Field, SecretStr
+from pydantic import BaseModel, Field, SecretStr, model_validator
 from starlette.status import HTTP_500_INTERNAL_SERVER_ERROR, HTTP_502_BAD_GATEWAY
 
 from backend.api.features.library.db import set_preset_webhook, update_preset
@@ -39,7 +39,11 @@ from backend.data.onboarding import OnboardingStep, complete_onboarding_step
 from backend.data.user import get_user_integrations
 from backend.executor.utils import add_graph_execution
 from backend.integrations.ayrshare import AyrshareClient, SocialPlatform
-from backend.integrations.creds_manager import IntegrationCredentialsManager
+from backend.integrations.credentials_store import provider_matches
+from backend.integrations.creds_manager import (
+    IntegrationCredentialsManager,
+    create_mcp_oauth_handler,
+)
 from backend.integrations.oauth import CREDENTIALS_BY_PROVIDER, HANDLERS_BY_NAME
 from backend.integrations.providers import ProviderName
 from backend.integrations.webhooks import get_webhook_manager
@@ -102,9 +106,37 @@ class CredentialsMetaResponse(BaseModel):
     scopes: list[str] | None
     username: str | None
     host: str | None = Field(
-        default=None, description="Host pattern for host-scoped credentials"
+        default=None,
+        description="Host pattern for host-scoped or MCP server URL for MCP credentials",
     )
 
+    @model_validator(mode="before")
+    @classmethod
+    def _normalize_provider(cls, data: Any) -> Any:
+        """Fix ``ProviderName.X`` format from Python 3.13 ``str(Enum)`` bug."""
+        if isinstance(data, dict):
+            prov = data.get("provider", "")
+            if isinstance(prov, str) and prov.startswith("ProviderName."):
+                member = prov.removeprefix("ProviderName.")
+                try:
+                    data = {**data, "provider": ProviderName[member].value}
+                except KeyError:
+                    pass
+        return data
+
+    @staticmethod
+    def get_host(cred: Credentials) -> str | None:
+        """Extract host from credential: HostScoped host or MCP server URL."""
+        if isinstance(cred, HostScopedCredentials):
+            return cred.host
+        if isinstance(cred, OAuth2Credentials) and cred.provider in (
+            ProviderName.MCP,
+            ProviderName.MCP.value,
+            "ProviderName.MCP",
+        ):
+            return (cred.metadata or {}).get("mcp_server_url")
+        return None
+
 
 @router.post("/{provider}/callback", summary="Exchange OAuth code for tokens")
 async def callback(
@@ -179,9 +211,7 @@ async def callback(
         title=credentials.title,
         scopes=credentials.scopes,
         username=credentials.username,
-        host=(
-            credentials.host if isinstance(credentials, HostScopedCredentials) else None
-        ),
+        host=(CredentialsMetaResponse.get_host(credentials)),
     )
 
 
@@ -199,7 +229,7 @@ async def list_credentials(
             title=cred.title,
             scopes=cred.scopes if isinstance(cred, OAuth2Credentials) else None,
             username=cred.username if isinstance(cred, OAuth2Credentials) else None,
-            host=cred.host if isinstance(cred, HostScopedCredentials) else None,
+            host=CredentialsMetaResponse.get_host(cred),
         )
         for cred in credentials
     ]
@@ -222,7 +252,7 @@ async def list_credentials_by_provider(
             title=cred.title,
             scopes=cred.scopes if isinstance(cred, OAuth2Credentials) else None,
             username=cred.username if isinstance(cred, OAuth2Credentials) else None,
-            host=cred.host if isinstance(cred, HostScopedCredentials) else None,
+            host=CredentialsMetaResponse.get_host(cred),
         )
         for cred in credentials
     ]
@@ -322,7 +352,11 @@ async def delete_credentials(
 
     tokens_revoked = None
     if isinstance(creds, OAuth2Credentials):
-        handler = _get_provider_oauth_handler(request, provider)
+        if provider_matches(provider.value, ProviderName.MCP.value):
+            # MCP uses dynamic per-server OAuth — create handler from metadata
+            handler = create_mcp_oauth_handler(creds)
+        else:
+            handler = _get_provider_oauth_handler(request, provider)
         tokens_revoked = await handler.revoke_tokens(creds)
 
     return CredentialsDeletionResponse(revoked=tokens_revoked)

autogpt_platform/backend/backend/api/features/mcp/routes.py

@@ -0,0 +1,404 @@
+"""
+MCP (Model Context Protocol) API routes.
+
+Provides endpoints for MCP tool discovery and OAuth authentication so the
+frontend can list available tools on an MCP server before placing a block.
+"""
+
+import logging
+from typing import Annotated, Any
+from urllib.parse import urlparse
+
+import fastapi
+from autogpt_libs.auth import get_user_id
+from fastapi import Security
+from pydantic import BaseModel, Field
+
+from backend.api.features.integrations.router import CredentialsMetaResponse
+from backend.blocks.mcp.client import MCPClient, MCPClientError
+from backend.blocks.mcp.oauth import MCPOAuthHandler
+from backend.data.model import OAuth2Credentials
+from backend.integrations.creds_manager import IntegrationCredentialsManager
+from backend.integrations.providers import ProviderName
+from backend.util.request import HTTPClientError, Requests
+from backend.util.settings import Settings
+
+logger = logging.getLogger(__name__)
+
+settings = Settings()
+router = fastapi.APIRouter(tags=["mcp"])
+creds_manager = IntegrationCredentialsManager()
+
+
+# ====================== Tool Discovery ====================== #
+
+
+class DiscoverToolsRequest(BaseModel):
+    """Request to discover tools on an MCP server."""
+
+    server_url: str = Field(description="URL of the MCP server")
+    auth_token: str | None = Field(
+        default=None,
+        description="Optional Bearer token for authenticated MCP servers",
+    )
+
+
+class MCPToolResponse(BaseModel):
+    """A single MCP tool returned by discovery."""
+
+    name: str
+    description: str
+    input_schema: dict[str, Any]
+
+
+class DiscoverToolsResponse(BaseModel):
+    """Response containing the list of tools available on an MCP server."""
+
+    tools: list[MCPToolResponse]
+    server_name: str | None = None
+    protocol_version: str | None = None
+
+
+@router.post(
+    "/discover-tools",
+    summary="Discover available tools on an MCP server",
+    response_model=DiscoverToolsResponse,
+)
+async def discover_tools(
+    request: DiscoverToolsRequest,
+    user_id: Annotated[str, Security(get_user_id)],
+) -> DiscoverToolsResponse:
+    """
+    Connect to an MCP server and return its available tools.
+
+    If the user has a stored MCP credential for this server URL, it will be
+    used automatically — no need to pass an explicit auth token.
+    """
+    auth_token = request.auth_token
+
+    # Auto-use stored MCP credential when no explicit token is provided.
+    if not auth_token:
+        mcp_creds = await creds_manager.store.get_creds_by_provider(
+            user_id, ProviderName.MCP.value
+        )
+        # Find the freshest credential for this server URL
+        best_cred: OAuth2Credentials | None = None
+        for cred in mcp_creds:
+            if (
+                isinstance(cred, OAuth2Credentials)
+                and (cred.metadata or {}).get("mcp_server_url") == request.server_url
+            ):
+                if best_cred is None or (
+                    (cred.access_token_expires_at or 0)
+                    > (best_cred.access_token_expires_at or 0)
+                ):
+                    best_cred = cred
+        if best_cred:
+            # Refresh the token if expired before using it
+            best_cred = await creds_manager.refresh_if_needed(user_id, best_cred)
+            logger.info(
+                f"Using MCP credential {best_cred.id} for {request.server_url}, "
+                f"expires_at={best_cred.access_token_expires_at}"
+            )
+            auth_token = best_cred.access_token.get_secret_value()
+
+    client = MCPClient(request.server_url, auth_token=auth_token)
+
+    try:
+        init_result = await client.initialize()
+        tools = await client.list_tools()
+    except HTTPClientError as e:
+        if e.status_code in (401, 403):
+            raise fastapi.HTTPException(
+                status_code=401,
+                detail="This MCP server requires authentication. "
+                "Please provide a valid auth token.",
+            )
+        raise fastapi.HTTPException(status_code=502, detail=str(e))
+    except MCPClientError as e:
+        raise fastapi.HTTPException(status_code=502, detail=str(e))
+    except Exception as e:
+        raise fastapi.HTTPException(
+            status_code=502,
+            detail=f"Failed to connect to MCP server: {e}",
+        )
+
+    return DiscoverToolsResponse(
+        tools=[
+            MCPToolResponse(
+                name=t.name,
+                description=t.description,
+                input_schema=t.input_schema,
+            )
+            for t in tools
+        ],
+        server_name=(
+            init_result.get("serverInfo", {}).get("name")
+            or urlparse(request.server_url).hostname
+            or "MCP"
+        ),
+        protocol_version=init_result.get("protocolVersion"),
+    )
+
+
+# ======================== OAuth Flow ======================== #
+
+
+class MCPOAuthLoginRequest(BaseModel):
+    """Request to start an OAuth flow for an MCP server."""
+
+    server_url: str = Field(description="URL of the MCP server that requires OAuth")
+
+
+class MCPOAuthLoginResponse(BaseModel):
+    """Response with the OAuth login URL for the user to authenticate."""
+
+    login_url: str
+    state_token: str
+
+
+@router.post(
+    "/oauth/login",
+    summary="Initiate OAuth login for an MCP server",
+)
+async def mcp_oauth_login(
+    request: MCPOAuthLoginRequest,
+    user_id: Annotated[str, Security(get_user_id)],
+) -> MCPOAuthLoginResponse:
+    """
+    Discover OAuth metadata from the MCP server and return a login URL.
+
+    1. Discovers the protected-resource metadata (RFC 9728)
+    2. Fetches the authorization server metadata (RFC 8414)
+    3. Performs Dynamic Client Registration (RFC 7591) if available
+    4. Returns the authorization URL for the frontend to open in a popup
+    """
+    client = MCPClient(request.server_url)
+
+    # Step 1: Discover protected-resource metadata (RFC 9728)
+    protected_resource = await client.discover_auth()
+
+    metadata: dict[str, Any] | None = None
+
+    if protected_resource and protected_resource.get("authorization_servers"):
+        auth_server_url = protected_resource["authorization_servers"][0]
+        resource_url = protected_resource.get("resource", request.server_url)
+
+        # Step 2a: Discover auth-server metadata (RFC 8414)
+        metadata = await client.discover_auth_server_metadata(auth_server_url)
+    else:
+        # Fallback: Some MCP servers (e.g. Linear) are their own auth server
+        # and serve OAuth metadata directly without protected-resource metadata.
+        # Don't assume a resource_url — omitting it lets the auth server choose
+        # the correct audience for the token (RFC 8707 resource is optional).
+        resource_url = None
+        metadata = await client.discover_auth_server_metadata(request.server_url)
+
+    if (
+        not metadata
+        or "authorization_endpoint" not in metadata
+        or "token_endpoint" not in metadata
+    ):
+        raise fastapi.HTTPException(
+            status_code=400,
+            detail="This MCP server does not advertise OAuth support. "
+            "You may need to provide an auth token manually.",
+        )
+
+    authorize_url = metadata["authorization_endpoint"]
+    token_url = metadata["token_endpoint"]
+    registration_endpoint = metadata.get("registration_endpoint")
+    revoke_url = metadata.get("revocation_endpoint")
+
+    # Step 3: Dynamic Client Registration (RFC 7591) if available
+    frontend_base_url = settings.config.frontend_base_url
+    if not frontend_base_url:
+        raise fastapi.HTTPException(
+            status_code=500,
+            detail="Frontend base URL is not configured.",
+        )
+    redirect_uri = f"{frontend_base_url}/auth/integrations/mcp_callback"
+
+    client_id = ""
+    client_secret = ""
+    if registration_endpoint:
+        reg_result = await _register_mcp_client(
+            registration_endpoint, redirect_uri, request.server_url
+        )
+        if reg_result:
+            client_id = reg_result.get("client_id", "")
+            client_secret = reg_result.get("client_secret", "")
+
+    if not client_id:
+        client_id = "autogpt-platform"
+
+    # Step 4: Store state token with OAuth metadata for the callback
+    scopes = (protected_resource or {}).get("scopes_supported") or metadata.get(
+        "scopes_supported", []
+    )
+    state_token, code_challenge = await creds_manager.store.store_state_token(
+        user_id,
+        ProviderName.MCP.value,
+        scopes,
+        state_metadata={
+            "authorize_url": authorize_url,
+            "token_url": token_url,
+            "revoke_url": revoke_url,
+            "resource_url": resource_url,
+            "server_url": request.server_url,
+            "client_id": client_id,
+            "client_secret": client_secret,
+        },
+    )
+
+    # Step 5: Build and return the login URL
+    handler = MCPOAuthHandler(
+        client_id=client_id,
+        client_secret=client_secret,
+        redirect_uri=redirect_uri,
+        authorize_url=authorize_url,
+        token_url=token_url,
+        resource_url=resource_url,
+    )
+    login_url = handler.get_login_url(
+        scopes, state_token, code_challenge=code_challenge
+    )
+
+    return MCPOAuthLoginResponse(login_url=login_url, state_token=state_token)
+
+
+class MCPOAuthCallbackRequest(BaseModel):
+    """Request to exchange an OAuth code for tokens."""
+
+    code: str = Field(description="Authorization code from OAuth callback")
+    state_token: str = Field(description="State token for CSRF verification")
+
+
+class MCPOAuthCallbackResponse(BaseModel):
+    """Response after successfully storing OAuth credentials."""
+
+    credential_id: str
+
+
+@router.post(
+    "/oauth/callback",
+    summary="Exchange OAuth code for MCP tokens",
+)
+async def mcp_oauth_callback(
+    request: MCPOAuthCallbackRequest,
+    user_id: Annotated[str, Security(get_user_id)],
+) -> CredentialsMetaResponse:
+    """
+    Exchange the authorization code for tokens and store the credential.
+
+    The frontend calls this after receiving the OAuth code from the popup.
+    On success, subsequent ``/discover-tools`` calls for the same server URL
+    will automatically use the stored credential.
+    """
+    valid_state = await creds_manager.store.verify_state_token(
+        user_id, request.state_token, ProviderName.MCP.value
+    )
+    if not valid_state:
+        raise fastapi.HTTPException(
+            status_code=400,
+            detail="Invalid or expired state token.",
+        )
+
+    meta = valid_state.state_metadata
+    frontend_base_url = settings.config.frontend_base_url
+    if not frontend_base_url:
+        raise fastapi.HTTPException(
+            status_code=500,
+            detail="Frontend base URL is not configured.",
+        )
+    redirect_uri = f"{frontend_base_url}/auth/integrations/mcp_callback"
+
+    handler = MCPOAuthHandler(
+        client_id=meta["client_id"],
+        client_secret=meta.get("client_secret", ""),
+        redirect_uri=redirect_uri,
+        authorize_url=meta["authorize_url"],
+        token_url=meta["token_url"],
+        revoke_url=meta.get("revoke_url"),
+        resource_url=meta.get("resource_url"),
+    )
+
+    try:
+        credentials = await handler.exchange_code_for_tokens(
+            request.code, valid_state.scopes, valid_state.code_verifier
+        )
+    except Exception as e:
+        raise fastapi.HTTPException(
+            status_code=400,
+            detail=f"OAuth token exchange failed: {e}",
+        )
+
+    # Enrich credential metadata for future lookup and token refresh
+    if credentials.metadata is None:
+        credentials.metadata = {}
+    credentials.metadata["mcp_server_url"] = meta["server_url"]
+    credentials.metadata["mcp_client_id"] = meta["client_id"]
+    credentials.metadata["mcp_client_secret"] = meta.get("client_secret", "")
+    credentials.metadata["mcp_token_url"] = meta["token_url"]
+    credentials.metadata["mcp_resource_url"] = meta.get("resource_url", "")
+
+    hostname = urlparse(meta["server_url"]).hostname or meta["server_url"]
+    credentials.title = f"MCP: {hostname}"
+
+    # Remove old MCP credentials for the same server to prevent stale token buildup.
+    try:
+        old_creds = await creds_manager.store.get_creds_by_provider(
+            user_id, ProviderName.MCP.value
+        )
+        for old in old_creds:
+            if (
+                isinstance(old, OAuth2Credentials)
+                and (old.metadata or {}).get("mcp_server_url") == meta["server_url"]
+            ):
+                await creds_manager.store.delete_creds_by_id(user_id, old.id)
+                logger.info(
+                    f"Removed old MCP credential {old.id} for {meta['server_url']}"
+                )
+    except Exception:
+        logger.debug("Could not clean up old MCP credentials", exc_info=True)
+
+    await creds_manager.create(user_id, credentials)
+
+    return CredentialsMetaResponse(
+        id=credentials.id,
+        provider=credentials.provider,
+        type=credentials.type,
+        title=credentials.title,
+        scopes=credentials.scopes,
+        username=credentials.username,
+        host=credentials.metadata.get("mcp_server_url"),
+    )
+
+
+# ======================== Helpers ======================== #
+
+
+async def _register_mcp_client(
+    registration_endpoint: str,
+    redirect_uri: str,
+    server_url: str,
+) -> dict[str, Any] | None:
+    """Attempt Dynamic Client Registration (RFC 7591) with an MCP auth server."""
+    try:
+        response = await Requests(raise_for_status=True).post(
+            registration_endpoint,
+            json={
+                "client_name": "AutoGPT Platform",
+                "redirect_uris": [redirect_uri],
+                "grant_types": ["authorization_code"],
+                "response_types": ["code"],
+                "token_endpoint_auth_method": "client_secret_post",
+            },
+        )
+        data = response.json()
+        if isinstance(data, dict) and "client_id" in data:
+            return data
+        return None
+    except Exception as e:
+        logger.warning(f"Dynamic client registration failed for {server_url}: {e}")
+        return None

autogpt_platform/backend/backend/api/features/mcp/test_routes.py

@@ -0,0 +1,436 @@
+"""Tests for MCP API routes.
+
+Uses httpx.AsyncClient with ASGITransport instead of fastapi.testclient.TestClient
+to avoid creating blocking portals that can corrupt pytest-asyncio's session event loop.
+"""
+
+from unittest.mock import AsyncMock, patch
+
+import fastapi
+import httpx
+import pytest
+import pytest_asyncio
+from autogpt_libs.auth import get_user_id
+
+from backend.api.features.mcp.routes import router
+from backend.blocks.mcp.client import MCPClientError, MCPTool
+from backend.util.request import HTTPClientError
+
+app = fastapi.FastAPI()
+app.include_router(router)
+app.dependency_overrides[get_user_id] = lambda: "test-user-id"
+
+
+@pytest_asyncio.fixture(scope="module")
+async def client():
+    transport = httpx.ASGITransport(app=app)
+    async with httpx.AsyncClient(transport=transport, base_url="http://test") as c:
+        yield c
+
+
+class TestDiscoverTools:
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_discover_tools_success(self, client):
+        mock_tools = [
+            MCPTool(
+                name="get_weather",
+                description="Get weather for a city",
+                input_schema={
+                    "type": "object",
+                    "properties": {"city": {"type": "string"}},
+                    "required": ["city"],
+                },
+            ),
+            MCPTool(
+                name="add_numbers",
+                description="Add two numbers",
+                input_schema={
+                    "type": "object",
+                    "properties": {
+                        "a": {"type": "number"},
+                        "b": {"type": "number"},
+                    },
+                },
+            ),
+        ]
+
+        with (
+            patch("backend.api.features.mcp.routes.MCPClient") as MockClient,
+            patch("backend.api.features.mcp.routes.creds_manager") as mock_cm,
+        ):
+            mock_cm.store.get_creds_by_provider = AsyncMock(return_value=[])
+            instance = MockClient.return_value
+            instance.initialize = AsyncMock(
+                return_value={
+                    "protocolVersion": "2025-03-26",
+                    "serverInfo": {"name": "test-server"},
+                }
+            )
+            instance.list_tools = AsyncMock(return_value=mock_tools)
+
+            response = await client.post(
+                "/discover-tools",
+                json={"server_url": "https://mcp.example.com/mcp"},
+            )
+
+        assert response.status_code == 200
+        data = response.json()
+        assert len(data["tools"]) == 2
+        assert data["tools"][0]["name"] == "get_weather"
+        assert data["tools"][1]["name"] == "add_numbers"
+        assert data["server_name"] == "test-server"
+        assert data["protocol_version"] == "2025-03-26"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_discover_tools_with_auth_token(self, client):
+        with patch("backend.api.features.mcp.routes.MCPClient") as MockClient:
+            instance = MockClient.return_value
+            instance.initialize = AsyncMock(
+                return_value={"serverInfo": {}, "protocolVersion": "2025-03-26"}
+            )
+            instance.list_tools = AsyncMock(return_value=[])
+
+            response = await client.post(
+                "/discover-tools",
+                json={
+                    "server_url": "https://mcp.example.com/mcp",
+                    "auth_token": "my-secret-token",
+                },
+            )
+
+        assert response.status_code == 200
+        MockClient.assert_called_once_with(
+            "https://mcp.example.com/mcp",
+            auth_token="my-secret-token",
+        )
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_discover_tools_auto_uses_stored_credential(self, client):
+        """When no explicit token is given, stored MCP credentials are used."""
+        from pydantic import SecretStr
+
+        from backend.data.model import OAuth2Credentials
+
+        stored_cred = OAuth2Credentials(
+            provider="mcp",
+            title="MCP: example.com",
+            access_token=SecretStr("stored-token-123"),
+            refresh_token=None,
+            access_token_expires_at=None,
+            refresh_token_expires_at=None,
+            scopes=[],
+            metadata={"mcp_server_url": "https://mcp.example.com/mcp"},
+        )
+
+        with (
+            patch("backend.api.features.mcp.routes.MCPClient") as MockClient,
+            patch("backend.api.features.mcp.routes.creds_manager") as mock_cm,
+        ):
+            mock_cm.store.get_creds_by_provider = AsyncMock(return_value=[stored_cred])
+            mock_cm.refresh_if_needed = AsyncMock(return_value=stored_cred)
+            instance = MockClient.return_value
+            instance.initialize = AsyncMock(
+                return_value={"serverInfo": {}, "protocolVersion": "2025-03-26"}
+            )
+            instance.list_tools = AsyncMock(return_value=[])
+
+            response = await client.post(
+                "/discover-tools",
+                json={"server_url": "https://mcp.example.com/mcp"},
+            )
+
+        assert response.status_code == 200
+        MockClient.assert_called_once_with(
+            "https://mcp.example.com/mcp",
+            auth_token="stored-token-123",
+        )
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_discover_tools_mcp_error(self, client):
+        with (
+            patch("backend.api.features.mcp.routes.MCPClient") as MockClient,
+            patch("backend.api.features.mcp.routes.creds_manager") as mock_cm,
+        ):
+            mock_cm.store.get_creds_by_provider = AsyncMock(return_value=[])
+            instance = MockClient.return_value
+            instance.initialize = AsyncMock(
+                side_effect=MCPClientError("Connection refused")
+            )
+
+            response = await client.post(
+                "/discover-tools",
+                json={"server_url": "https://bad-server.example.com/mcp"},
+            )
+
+        assert response.status_code == 502
+        assert "Connection refused" in response.json()["detail"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_discover_tools_generic_error(self, client):
+        with (
+            patch("backend.api.features.mcp.routes.MCPClient") as MockClient,
+            patch("backend.api.features.mcp.routes.creds_manager") as mock_cm,
+        ):
+            mock_cm.store.get_creds_by_provider = AsyncMock(return_value=[])
+            instance = MockClient.return_value
+            instance.initialize = AsyncMock(side_effect=Exception("Network timeout"))
+
+            response = await client.post(
+                "/discover-tools",
+                json={"server_url": "https://timeout.example.com/mcp"},
+            )
+
+        assert response.status_code == 502
+        assert "Failed to connect" in response.json()["detail"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_discover_tools_auth_required(self, client):
+        with (
+            patch("backend.api.features.mcp.routes.MCPClient") as MockClient,
+            patch("backend.api.features.mcp.routes.creds_manager") as mock_cm,
+        ):
+            mock_cm.store.get_creds_by_provider = AsyncMock(return_value=[])
+            instance = MockClient.return_value
+            instance.initialize = AsyncMock(
+                side_effect=HTTPClientError("HTTP 401 Error: Unauthorized", 401)
+            )
+
+            response = await client.post(
+                "/discover-tools",
+                json={"server_url": "https://auth-server.example.com/mcp"},
+            )
+
+        assert response.status_code == 401
+        assert "requires authentication" in response.json()["detail"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_discover_tools_forbidden(self, client):
+        with (
+            patch("backend.api.features.mcp.routes.MCPClient") as MockClient,
+            patch("backend.api.features.mcp.routes.creds_manager") as mock_cm,
+        ):
+            mock_cm.store.get_creds_by_provider = AsyncMock(return_value=[])
+            instance = MockClient.return_value
+            instance.initialize = AsyncMock(
+                side_effect=HTTPClientError("HTTP 403 Error: Forbidden", 403)
+            )
+
+            response = await client.post(
+                "/discover-tools",
+                json={"server_url": "https://auth-server.example.com/mcp"},
+            )
+
+        assert response.status_code == 401
+        assert "requires authentication" in response.json()["detail"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_discover_tools_missing_url(self, client):
+        response = await client.post("/discover-tools", json={})
+        assert response.status_code == 422
+
+
+class TestOAuthLogin:
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_oauth_login_success(self, client):
+        with (
+            patch("backend.api.features.mcp.routes.MCPClient") as MockClient,
+            patch("backend.api.features.mcp.routes.creds_manager") as mock_cm,
+            patch("backend.api.features.mcp.routes.settings") as mock_settings,
+            patch(
+                "backend.api.features.mcp.routes._register_mcp_client"
+            ) as mock_register,
+        ):
+            instance = MockClient.return_value
+            instance.discover_auth = AsyncMock(
+                return_value={
+                    "authorization_servers": ["https://auth.sentry.io"],
+                    "resource": "https://mcp.sentry.dev/mcp",
+                    "scopes_supported": ["openid"],
+                }
+            )
+            instance.discover_auth_server_metadata = AsyncMock(
+                return_value={
+                    "authorization_endpoint": "https://auth.sentry.io/authorize",
+                    "token_endpoint": "https://auth.sentry.io/token",
+                    "registration_endpoint": "https://auth.sentry.io/register",
+                }
+            )
+            mock_register.return_value = {
+                "client_id": "registered-client-id",
+                "client_secret": "registered-secret",
+            }
+            mock_cm.store.store_state_token = AsyncMock(
+                return_value=("state-token-123", "code-challenge-abc")
+            )
+            mock_settings.config.frontend_base_url = "http://localhost:3000"
+
+            response = await client.post(
+                "/oauth/login",
+                json={"server_url": "https://mcp.sentry.dev/mcp"},
+            )
+
+        assert response.status_code == 200
+        data = response.json()
+        assert "login_url" in data
+        assert data["state_token"] == "state-token-123"
+        assert "auth.sentry.io/authorize" in data["login_url"]
+        assert "registered-client-id" in data["login_url"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_oauth_login_no_oauth_support(self, client):
+        with patch("backend.api.features.mcp.routes.MCPClient") as MockClient:
+            instance = MockClient.return_value
+            instance.discover_auth = AsyncMock(return_value=None)
+            instance.discover_auth_server_metadata = AsyncMock(return_value=None)
+
+            response = await client.post(
+                "/oauth/login",
+                json={"server_url": "https://simple-server.example.com/mcp"},
+            )
+
+        assert response.status_code == 400
+        assert "does not advertise OAuth" in response.json()["detail"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_oauth_login_fallback_to_public_client(self, client):
+        """When DCR is unavailable, falls back to default public client ID."""
+        with (
+            patch("backend.api.features.mcp.routes.MCPClient") as MockClient,
+            patch("backend.api.features.mcp.routes.creds_manager") as mock_cm,
+            patch("backend.api.features.mcp.routes.settings") as mock_settings,
+        ):
+            instance = MockClient.return_value
+            instance.discover_auth = AsyncMock(
+                return_value={
+                    "authorization_servers": ["https://auth.example.com"],
+                    "resource": "https://mcp.example.com/mcp",
+                }
+            )
+            instance.discover_auth_server_metadata = AsyncMock(
+                return_value={
+                    "authorization_endpoint": "https://auth.example.com/authorize",
+                    "token_endpoint": "https://auth.example.com/token",
+                    # No registration_endpoint
+                }
+            )
+            mock_cm.store.store_state_token = AsyncMock(
+                return_value=("state-abc", "challenge-xyz")
+            )
+            mock_settings.config.frontend_base_url = "http://localhost:3000"
+
+            response = await client.post(
+                "/oauth/login",
+                json={"server_url": "https://mcp.example.com/mcp"},
+            )
+
+        assert response.status_code == 200
+        data = response.json()
+        assert "autogpt-platform" in data["login_url"]
+
+
+class TestOAuthCallback:
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_oauth_callback_success(self, client):
+        from pydantic import SecretStr
+
+        from backend.data.model import OAuth2Credentials
+
+        mock_creds = OAuth2Credentials(
+            provider="mcp",
+            title=None,
+            access_token=SecretStr("access-token-xyz"),
+            refresh_token=None,
+            access_token_expires_at=None,
+            refresh_token_expires_at=None,
+            scopes=[],
+            metadata={
+                "mcp_token_url": "https://auth.sentry.io/token",
+                "mcp_resource_url": "https://mcp.sentry.dev/mcp",
+            },
+        )
+
+        with (
+            patch("backend.api.features.mcp.routes.creds_manager") as mock_cm,
+            patch("backend.api.features.mcp.routes.settings") as mock_settings,
+            patch("backend.api.features.mcp.routes.MCPOAuthHandler") as MockHandler,
+        ):
+            mock_settings.config.frontend_base_url = "http://localhost:3000"
+
+            # Mock state verification
+            mock_state = AsyncMock()
+            mock_state.state_metadata = {
+                "authorize_url": "https://auth.sentry.io/authorize",
+                "token_url": "https://auth.sentry.io/token",
+                "client_id": "test-client-id",
+                "client_secret": "test-secret",
+                "server_url": "https://mcp.sentry.dev/mcp",
+            }
+            mock_state.scopes = ["openid"]
+            mock_state.code_verifier = "verifier-123"
+            mock_cm.store.verify_state_token = AsyncMock(return_value=mock_state)
+            mock_cm.create = AsyncMock()
+
+            handler_instance = MockHandler.return_value
+            handler_instance.exchange_code_for_tokens = AsyncMock(
+                return_value=mock_creds
+            )
+
+            # Mock old credential cleanup
+            mock_cm.store.get_creds_by_provider = AsyncMock(return_value=[])
+
+            response = await client.post(
+                "/oauth/callback",
+                json={"code": "auth-code-abc", "state_token": "state-token-123"},
+            )
+
+        assert response.status_code == 200
+        data = response.json()
+        assert "id" in data
+        assert data["provider"] == "mcp"
+        assert data["type"] == "oauth2"
+        mock_cm.create.assert_called_once()
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_oauth_callback_invalid_state(self, client):
+        with patch("backend.api.features.mcp.routes.creds_manager") as mock_cm:
+            mock_cm.store.verify_state_token = AsyncMock(return_value=None)
+
+            response = await client.post(
+                "/oauth/callback",
+                json={"code": "auth-code", "state_token": "bad-state"},
+            )
+
+        assert response.status_code == 400
+        assert "Invalid or expired" in response.json()["detail"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_oauth_callback_token_exchange_fails(self, client):
+        with (
+            patch("backend.api.features.mcp.routes.creds_manager") as mock_cm,
+            patch("backend.api.features.mcp.routes.settings") as mock_settings,
+            patch("backend.api.features.mcp.routes.MCPOAuthHandler") as MockHandler,
+        ):
+            mock_settings.config.frontend_base_url = "http://localhost:3000"
+            mock_state = AsyncMock()
+            mock_state.state_metadata = {
+                "authorize_url": "https://auth.example.com/authorize",
+                "token_url": "https://auth.example.com/token",
+                "client_id": "cid",
+                "server_url": "https://mcp.example.com/mcp",
+            }
+            mock_state.scopes = []
+            mock_state.code_verifier = "v"
+            mock_cm.store.verify_state_token = AsyncMock(return_value=mock_state)
+
+            handler_instance = MockHandler.return_value
+            handler_instance.exchange_code_for_tokens = AsyncMock(
+                side_effect=RuntimeError("Token exchange failed")
+            )
+
+            response = await client.post(
+                "/oauth/callback",
+                json={"code": "bad-code", "state_token": "state"},
+            )
+
+        assert response.status_code == 400
+        assert "token exchange failed" in response.json()["detail"].lower()

autogpt_platform/backend/backend/api/features/store/routes.py

@@ -393,7 +393,6 @@ async def get_creators(
 @router.get(
     "/creator/{username}",
     summary="Get creator details",
-    operation_id="getV2GetCreatorDetails",
     tags=["store", "public"],
     response_model=store_model.CreatorDetails,
 )

autogpt_platform/backend/backend/api/rest_api.py

@@ -18,7 +18,6 @@ from prisma.errors import PrismaError
 
 import backend.api.features.admin.credit_admin_routes
 import backend.api.features.admin.execution_analytics_routes
-import backend.api.features.admin.llm_routes
 import backend.api.features.admin.store_admin_routes
 import backend.api.features.builder
 import backend.api.features.builder.routes
@@ -27,6 +26,7 @@ import backend.api.features.executions.review.routes
 import backend.api.features.library.db
 import backend.api.features.library.model
 import backend.api.features.library.routes
+import backend.api.features.mcp.routes as mcp_routes
 import backend.api.features.oauth
 import backend.api.features.otto.routes
 import backend.api.features.postmark.postmark
@@ -39,15 +39,13 @@ import backend.data.db
 import backend.data.graph
 import backend.data.user
 import backend.integrations.webhooks.utils
-import backend.server.v2.llm.routes as public_llm_routes
 import backend.util.service
 import backend.util.settings
 from backend.api.features.chat.completion_consumer import (
     start_completion_consumer,
     stop_completion_consumer,
 )
-from backend.data import llm_registry
-from backend.data.block_cost_config import refresh_llm_costs
+from backend.blocks.llm import DEFAULT_LLM_MODEL
 from backend.data.model import Credentials
 from backend.integrations.providers import ProviderName
 from backend.monitoring.instrumentation import instrument_fastapi
@@ -118,27 +116,11 @@ async def lifespan_context(app: fastapi.FastAPI):
 
     AutoRegistry.patch_integrations()
 
-    # Refresh LLM registry before initializing blocks so blocks can use registry data
-    await llm_registry.refresh_llm_registry()
-    await refresh_llm_costs()
-
-    # Clear block schema caches so they're regenerated with updated discriminator_mapping
-    from backend.blocks._base import BlockSchema
-
-    BlockSchema.clear_all_schema_caches()
-
     await backend.data.block.initialize_blocks()
 
     await backend.data.user.migrate_and_encrypt_user_integrations()
     await backend.data.graph.fix_llm_provider_credentials()
-    # migrate_llm_models uses registry default model
-    from backend.blocks.llm import LlmModel
-
-    default_model_slug = llm_registry.get_default_model_slug()
-    if default_model_slug:
-        await backend.data.graph.migrate_llm_models(LlmModel(default_model_slug))
-    else:
-        logger.warning("Skipping LLM model migration: no default model available")
+    await backend.data.graph.migrate_llm_models(DEFAULT_LLM_MODEL)
     await backend.integrations.webhooks.utils.migrate_legacy_triggered_graphs()
 
     # Start chat completion consumer for Redis Streams notifications
@@ -340,16 +322,6 @@ app.include_router(
     tags=["v2", "executions", "review"],
     prefix="/api/review",
 )
-app.include_router(
-    backend.api.features.admin.llm_routes.router,
-    tags=["v2", "admin", "llm"],
-    prefix="/api/llm/admin",
-)
-app.include_router(
-    public_llm_routes.router,
-    tags=["v2", "llm"],
-    prefix="/api",
-)
 app.include_router(
     backend.api.features.library.routes.router, tags=["v2"], prefix="/api/library"
 )
@@ -372,6 +344,11 @@ app.include_router(
     tags=["workspace"],
     prefix="/api/workspace",
 )
+app.include_router(
+    mcp_routes.router,
+    tags=["v2", "mcp"],
+    prefix="/api/mcp",
+)
 app.include_router(
     backend.api.features.oauth.router,
     tags=["oauth"],

autogpt_platform/backend/backend/api/ws_api.py

@@ -79,49 +79,11 @@ async def event_broadcaster(manager: ConnectionManager):
                     payload=notification.payload,
                 )
 
-        # Track registry pubsub for cleanup
-        registry_pubsub = None
-
-        async def registry_refresh_worker():
-            """Listen for LLM registry refresh notifications and broadcast to all clients."""
-            nonlocal registry_pubsub
-            from backend.data.llm_registry import REGISTRY_REFRESH_CHANNEL
-            from backend.data.redis_client import connect_async
-
-            redis = await connect_async()
-            registry_pubsub = redis.pubsub()
-            await registry_pubsub.subscribe(REGISTRY_REFRESH_CHANNEL)
-            logger.info(
-                "Subscribed to LLM registry refresh notifications for WebSocket broadcast"
-            )
-
-            async for message in registry_pubsub.listen():
-                if (
-                    message["type"] == "message"
-                    and message["channel"] == REGISTRY_REFRESH_CHANNEL
-                ):
-                    logger.info(
-                        "Broadcasting LLM registry refresh to all WebSocket clients"
-                    )
-                    await manager.broadcast_to_all(
-                        method=WSMethod.NOTIFICATION,
-                        data={
-                            "type": "LLM_REGISTRY_REFRESH",
-                            "event": "registry_updated",
-                        },
-                    )
-
-        await asyncio.gather(
-            execution_worker(),
-            notification_worker(),
-            registry_refresh_worker(),
-        )
+        await asyncio.gather(execution_worker(), notification_worker())
     finally:
         # Ensure PubSub connections are closed on any exit to prevent leaks
         await execution_bus.close()
         await notification_bus.close()
-        if registry_pubsub:
-            await registry_pubsub.close()
 
 
 async def authenticate_websocket(websocket: WebSocket) -> str:

autogpt_platform/backend/backend/blocks/_base.py

@@ -64,6 +64,7 @@ class BlockType(Enum):
     AI = "AI"
     AYRSHARE = "Ayrshare"
     HUMAN_IN_THE_LOOP = "Human In The Loop"
+    MCP_TOOL = "MCP Tool"
 
 
 class BlockCategory(Enum):
@@ -133,26 +134,7 @@ class BlockInfo(BaseModel):
 
 
 class BlockSchema(BaseModel):
-    cached_jsonschema: ClassVar[dict[str, Any] | None] = None
-
-    @classmethod
-    def clear_schema_cache(cls) -> None:
-        """Clear the cached JSON schema for this class."""
-        # Use None instead of {} because {} is truthy and would prevent regeneration
-        cls.cached_jsonschema = None  # type: ignore
-
-    @staticmethod
-    def clear_all_schema_caches() -> None:
-        """Clear cached JSON schemas for all BlockSchema subclasses."""
-
-        def clear_recursive(cls: type) -> None:
-            """Recursively clear cache for class and all subclasses."""
-            if hasattr(cls, "clear_schema_cache"):
-                cls.clear_schema_cache()
-            for subclass in cls.__subclasses__():
-                clear_recursive(subclass)
-
-        clear_recursive(BlockSchema)
+    cached_jsonschema: ClassVar[dict[str, Any]]
 
     @classmethod
     def jsonschema(cls) -> dict[str, Any]:
@@ -243,8 +225,7 @@ class BlockSchema(BaseModel):
         super().__pydantic_init_subclass__(**kwargs)
 
         # Reset cached JSON schema to prevent inheriting it from parent class
-        # Use None instead of {} because {} is truthy and would prevent regeneration
-        cls.cached_jsonschema = None
+        cls.cached_jsonschema = {}
 
         credentials_fields = cls.get_credentials_fields()
 

autogpt_platform/backend/backend/blocks/ai_condition.py

@@ -7,6 +7,7 @@ from backend.blocks._base import (
     BlockSchemaOutput,
 )
 from backend.blocks.llm import (
+    DEFAULT_LLM_MODEL,
     TEST_CREDENTIALS,
     TEST_CREDENTIALS_INPUT,
     AIBlockBase,
@@ -15,7 +16,6 @@ from backend.blocks.llm import (
     LlmModel,
     LLMResponse,
     llm_call,
-    llm_model_schema_extra,
 )
 from backend.data.model import APIKeyCredentials, NodeExecutionStats, SchemaField
 
@@ -50,10 +50,9 @@ class AIConditionBlock(AIBlockBase):
         )
         model: LlmModel = SchemaField(
             title="LLM Model",
-            default_factory=LlmModel.default,
+            default=DEFAULT_LLM_MODEL,
             description="The language model to use for evaluating the condition.",
             advanced=False,
-            json_schema_extra=llm_model_schema_extra(),
         )
         credentials: AICredentials = AICredentialsField()
 
@@ -83,7 +82,7 @@ class AIConditionBlock(AIBlockBase):
                 "condition": "the input is an email address",
                 "yes_value": "Valid email",
                 "no_value": "Not an email",
-                "model": LlmModel.default(),
+                "model": DEFAULT_LLM_MODEL,
                 "credentials": TEST_CREDENTIALS_INPUT,
             },
             test_credentials=TEST_CREDENTIALS,

autogpt_platform/backend/backend/blocks/basic.py

@@ -126,6 +126,7 @@ class PrintToConsoleBlock(Block):
             output_schema=PrintToConsoleBlock.Output,
             test_input={"text": "Hello, World!"},
             is_sensitive_action=True,
+            disabled=True,  # Disabled per Nick Tindle's request (OPEN-3000)
             test_output=[
                 ("output", "Hello, World!"),
                 ("status", "printed"),

autogpt_platform/backend/backend/blocks/jina/search.py

@@ -17,6 +17,7 @@ from backend.blocks.jina._auth import (
 from backend.blocks.search import GetRequest
 from backend.data.model import SchemaField
 from backend.util.exceptions import BlockExecutionError
+from backend.util.request import HTTPClientError, HTTPServerError, validate_url
 
 
 class SearchTheWebBlock(Block, GetRequest):
@@ -110,7 +111,12 @@ class ExtractWebsiteContentBlock(Block, GetRequest):
         self, input_data: Input, *, credentials: JinaCredentials, **kwargs
     ) -> BlockOutput:
         if input_data.raw_content:
-            url = input_data.url
+            try:
+                parsed_url, _, _ = await validate_url(input_data.url, [])
+                url = parsed_url.geturl()
+            except ValueError as e:
+                yield "error", f"Invalid URL: {e}"
+                return
             headers = {}
         else:
             url = f"https://r.jina.ai/{input_data.url}"
@@ -119,5 +125,20 @@ class ExtractWebsiteContentBlock(Block, GetRequest):
                 "Authorization": f"Bearer {credentials.api_key.get_secret_value()}",
             }
 
-        content = await self.get_request(url, json=False, headers=headers)
+        try:
+            content = await self.get_request(url, json=False, headers=headers)
+        except HTTPClientError as e:
+            yield "error", f"Client error ({e.status_code}) fetching {input_data.url}: {e}"
+            return
+        except HTTPServerError as e:
+            yield "error", f"Server error ({e.status_code}) fetching {input_data.url}: {e}"
+            return
+        except Exception as e:
+            yield "error", f"Failed to fetch {input_data.url}: {e}"
+            return
+
+        if not content:
+            yield "error", f"No content returned for {input_data.url}"
+            return
+
         yield "content", content

autogpt_platform/backend/backend/blocks/llm.py

@@ -4,18 +4,16 @@ import logging
 import re
 import secrets
 from abc import ABC
-from dataclasses import dataclass
-from enum import Enum
+from enum import Enum, EnumMeta
 from json import JSONDecodeError
-from typing import Any, Iterable, List, Literal, Optional
+from typing import Any, Iterable, List, Literal, NamedTuple, Optional
 
 import anthropic
 import ollama
 import openai
 from anthropic.types import ToolParam
 from groq import AsyncGroq
-from pydantic import BaseModel, GetCoreSchemaHandler, SecretStr
-from pydantic_core import CoreSchema, core_schema
+from pydantic import BaseModel, SecretStr
 
 from backend.blocks._base import (
     Block,
@@ -24,8 +22,6 @@ from backend.blocks._base import (
     BlockSchemaInput,
     BlockSchemaOutput,
 )
-from backend.data import llm_registry
-from backend.data.llm_registry import ModelMetadata
 from backend.data.model import (
     APIKeyCredentials,
     CredentialsField,
@@ -70,123 +66,114 @@ TEST_CREDENTIALS_INPUT = {
 
 
 def AICredentialsField() -> AICredentials:
-    """
-    Returns a CredentialsField for LLM providers.
-    The discriminator_mapping will be refreshed when the schema is generated
-    if it's empty, ensuring the LLM registry is loaded.
-    """
-    # Get the mapping now - it may be empty initially, but will be refreshed
-    # when the schema is generated via CredentialsMetaInput._add_json_schema_extra
-    mapping = llm_registry.get_llm_discriminator_mapping()
-
     return CredentialsField(
         description="API key for the LLM provider.",
         discriminator="model",
-        discriminator_mapping=mapping,  # May be empty initially, refreshed later
+        discriminator_mapping={
+            model.value: model.metadata.provider for model in LlmModel
+        },
     )
 
 
-def llm_model_schema_extra() -> dict[str, Any]:
-    return {"options": llm_registry.get_llm_model_schema_options()}
+class ModelMetadata(NamedTuple):
+    provider: str
+    context_window: int
+    max_output_tokens: int | None
+    display_name: str
+    provider_name: str
+    creator_name: str
+    price_tier: Literal[1, 2, 3]
 
 
-class LlmModelMeta(type):
-    """
-    Metaclass for LlmModel that enables attribute-style access to dynamic models.
-
-    This allows code like `LlmModel.GPT4O` to work by converting the attribute
-    name to a slug format:
-    - GPT4O -> gpt-4o
-    - GPT4O_MINI -> gpt-4o-mini
-    - CLAUDE_3_5_SONNET -> claude-3-5-sonnet
-    """
-
-    def __getattr__(cls, name: str):
-        # Don't intercept private/dunder attributes
-        if name.startswith("_"):
-            raise AttributeError(f"type object 'LlmModel' has no attribute '{name}'")
-
-        # Convert attribute name to slug format:
-        # 1. Lowercase: GPT4O -> gpt4o
-        # 2. Underscores to hyphens: GPT4O_MINI -> gpt4o-mini
-        slug = name.lower().replace("_", "-")
-
-        # Check for exact match in registry first (e.g., "o1" stays "o1")
-        registry_slugs = llm_registry.get_dynamic_model_slugs()
-        if slug in registry_slugs:
-            return cls(slug)
-
-        # If no exact match, try inserting hyphen between letter and digit
-        # e.g., gpt4o -> gpt-4o
-        transformed_slug = re.sub(r"([a-z])(\d)", r"\1-\2", slug)
-        return cls(transformed_slug)
-
-    def __iter__(cls):
-        """Iterate over all models from the registry.
-
-        Yields LlmModel instances for each model in the dynamic registry.
-        Used by __get_pydantic_json_schema__ to build model metadata.
-        """
-        for model in llm_registry.iter_dynamic_models():
-            yield cls(model.slug)
+class LlmModelMeta(EnumMeta):
+    pass
 
 
-class LlmModel(str, metaclass=LlmModelMeta):
-    """
-    Dynamic LLM model type that accepts any model slug from the registry.
-
-    This is a string subclass (not an Enum) that allows any model slug value.
-    All models are managed via the LLM Registry in the database.
-
-    Usage:
-        model = LlmModel("gpt-4o")  # Direct construction
-        model = LlmModel.GPT4O      # Attribute access (converted to "gpt-4o")
-        model.value                  # Returns the slug string
-        model.provider               # Returns the provider from registry
-    """
-
-    def __new__(cls, value: str):
-        if isinstance(value, LlmModel):
-            return value
-        return str.__new__(cls, value)
-
-    @classmethod
-    def __get_pydantic_core_schema__(
-        cls, source_type: Any, handler: GetCoreSchemaHandler
-    ) -> CoreSchema:
-        """
-        Tell Pydantic how to validate LlmModel.
-
-        Accepts strings and converts them to LlmModel instances.
-        """
-        return core_schema.no_info_after_validator_function(
-            cls,  # The validator function (LlmModel constructor)
-            core_schema.str_schema(),  # Accept string input
-            serialization=core_schema.to_string_ser_schema(),  # Serialize as string
-        )
-
-    @property
-    def value(self) -> str:
-        """Return the model slug (for compatibility with enum-style access)."""
-        return str(self)
-
-    @classmethod
-    def default(cls) -> "LlmModel":
-        """
-        Get the default model from the registry.
-
-        Returns the recommended model if set, otherwise gpt-4o if available
-        and enabled, otherwise the first enabled model from the registry.
-        Falls back to "gpt-4o" if registry is empty (e.g., at module import time).
-        """
-        from backend.data.llm_registry import get_default_model_slug
-
-        slug = get_default_model_slug()
-        if slug is None:
-            # Registry is empty (e.g., at module import time before DB connection).
-            # Fall back to gpt-4o for backward compatibility.
-            slug = "gpt-4o"
-        return cls(slug)
+class LlmModel(str, Enum, metaclass=LlmModelMeta):
+    # OpenAI models
+    O3_MINI = "o3-mini"
+    O3 = "o3-2025-04-16"
+    O1 = "o1"
+    O1_MINI = "o1-mini"
+    # GPT-5 models
+    GPT5_2 = "gpt-5.2-2025-12-11"
+    GPT5_1 = "gpt-5.1-2025-11-13"
+    GPT5 = "gpt-5-2025-08-07"
+    GPT5_MINI = "gpt-5-mini-2025-08-07"
+    GPT5_NANO = "gpt-5-nano-2025-08-07"
+    GPT5_CHAT = "gpt-5-chat-latest"
+    GPT41 = "gpt-4.1-2025-04-14"
+    GPT41_MINI = "gpt-4.1-mini-2025-04-14"
+    GPT4O_MINI = "gpt-4o-mini"
+    GPT4O = "gpt-4o"
+    GPT4_TURBO = "gpt-4-turbo"
+    GPT3_5_TURBO = "gpt-3.5-turbo"
+    # Anthropic models
+    CLAUDE_4_1_OPUS = "claude-opus-4-1-20250805"
+    CLAUDE_4_OPUS = "claude-opus-4-20250514"
+    CLAUDE_4_SONNET = "claude-sonnet-4-20250514"
+    CLAUDE_4_5_OPUS = "claude-opus-4-5-20251101"
+    CLAUDE_4_5_SONNET = "claude-sonnet-4-5-20250929"
+    CLAUDE_4_5_HAIKU = "claude-haiku-4-5-20251001"
+    CLAUDE_4_6_OPUS = "claude-opus-4-6"
+    CLAUDE_3_HAIKU = "claude-3-haiku-20240307"
+    # AI/ML API models
+    AIML_API_QWEN2_5_72B = "Qwen/Qwen2.5-72B-Instruct-Turbo"
+    AIML_API_LLAMA3_1_70B = "nvidia/llama-3.1-nemotron-70b-instruct"
+    AIML_API_LLAMA3_3_70B = "meta-llama/Llama-3.3-70B-Instruct-Turbo"
+    AIML_API_META_LLAMA_3_1_70B = "meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo"
+    AIML_API_LLAMA_3_2_3B = "meta-llama/Llama-3.2-3B-Instruct-Turbo"
+    # Groq models
+    LLAMA3_3_70B = "llama-3.3-70b-versatile"
+    LLAMA3_1_8B = "llama-3.1-8b-instant"
+    # Ollama models
+    OLLAMA_LLAMA3_3 = "llama3.3"
+    OLLAMA_LLAMA3_2 = "llama3.2"
+    OLLAMA_LLAMA3_8B = "llama3"
+    OLLAMA_LLAMA3_405B = "llama3.1:405b"
+    OLLAMA_DOLPHIN = "dolphin-mistral:latest"
+    # OpenRouter models
+    OPENAI_GPT_OSS_120B = "openai/gpt-oss-120b"
+    OPENAI_GPT_OSS_20B = "openai/gpt-oss-20b"
+    GEMINI_2_5_PRO = "google/gemini-2.5-pro-preview-03-25"
+    GEMINI_3_PRO_PREVIEW = "google/gemini-3-pro-preview"
+    GEMINI_2_5_FLASH = "google/gemini-2.5-flash"
+    GEMINI_2_0_FLASH = "google/gemini-2.0-flash-001"
+    GEMINI_2_5_FLASH_LITE_PREVIEW = "google/gemini-2.5-flash-lite-preview-06-17"
+    GEMINI_2_0_FLASH_LITE = "google/gemini-2.0-flash-lite-001"
+    MISTRAL_NEMO = "mistralai/mistral-nemo"
+    COHERE_COMMAND_R_08_2024 = "cohere/command-r-08-2024"
+    COHERE_COMMAND_R_PLUS_08_2024 = "cohere/command-r-plus-08-2024"
+    DEEPSEEK_CHAT = "deepseek/deepseek-chat"  # Actually: DeepSeek V3
+    DEEPSEEK_R1_0528 = "deepseek/deepseek-r1-0528"
+    PERPLEXITY_SONAR = "perplexity/sonar"
+    PERPLEXITY_SONAR_PRO = "perplexity/sonar-pro"
+    PERPLEXITY_SONAR_DEEP_RESEARCH = "perplexity/sonar-deep-research"
+    NOUSRESEARCH_HERMES_3_LLAMA_3_1_405B = "nousresearch/hermes-3-llama-3.1-405b"
+    NOUSRESEARCH_HERMES_3_LLAMA_3_1_70B = "nousresearch/hermes-3-llama-3.1-70b"
+    AMAZON_NOVA_LITE_V1 = "amazon/nova-lite-v1"
+    AMAZON_NOVA_MICRO_V1 = "amazon/nova-micro-v1"
+    AMAZON_NOVA_PRO_V1 = "amazon/nova-pro-v1"
+    MICROSOFT_WIZARDLM_2_8X22B = "microsoft/wizardlm-2-8x22b"
+    GRYPHE_MYTHOMAX_L2_13B = "gryphe/mythomax-l2-13b"
+    META_LLAMA_4_SCOUT = "meta-llama/llama-4-scout"
+    META_LLAMA_4_MAVERICK = "meta-llama/llama-4-maverick"
+    GROK_4 = "x-ai/grok-4"
+    GROK_4_FAST = "x-ai/grok-4-fast"
+    GROK_4_1_FAST = "x-ai/grok-4.1-fast"
+    GROK_CODE_FAST_1 = "x-ai/grok-code-fast-1"
+    KIMI_K2 = "moonshotai/kimi-k2"
+    QWEN3_235B_A22B_THINKING = "qwen/qwen3-235b-a22b-thinking-2507"
+    QWEN3_CODER = "qwen/qwen3-coder"
+    # Llama API models
+    LLAMA_API_LLAMA_4_SCOUT = "Llama-4-Scout-17B-16E-Instruct-FP8"
+    LLAMA_API_LLAMA4_MAVERICK = "Llama-4-Maverick-17B-128E-Instruct-FP8"
+    LLAMA_API_LLAMA3_3_8B = "Llama-3.3-8B-Instruct"
+    LLAMA_API_LLAMA3_3_70B = "Llama-3.3-70B-Instruct"
+    # v0 by Vercel models
+    V0_1_5_MD = "v0-1.5-md"
+    V0_1_5_LG = "v0-1.5-lg"
+    V0_1_0_MD = "v0-1.0-md"
 
     @classmethod
     def __get_pydantic_json_schema__(cls, schema, handler):
@@ -194,15 +181,7 @@ class LlmModel(str, metaclass=LlmModelMeta):
         llm_model_metadata = {}
         for model in cls:
             model_name = model.value
-            # Skip disabled models - only show enabled models in the picker
-            if not llm_registry.is_model_enabled(model_name):
-                continue
-            # Use registry directly with None check to gracefully handle
-            # missing metadata during startup/import before registry is populated
-            metadata = llm_registry.get_llm_model_metadata(model_name)
-            if metadata is None:
-                # Skip models without metadata (registry not yet populated)
-                continue
+            metadata = model.metadata
             llm_model_metadata[model_name] = {
                 "creator": metadata.creator_name,
                 "creator_name": metadata.creator_name,
@@ -218,12 +197,7 @@ class LlmModel(str, metaclass=LlmModelMeta):
 
     @property
     def metadata(self) -> ModelMetadata:
-        metadata = llm_registry.get_llm_model_metadata(self.value)
-        if metadata:
-            return metadata
-        raise ValueError(
-            f"Missing metadata for model: {self.value}. Model not found in LLM registry."
-        )
+        return MODEL_METADATA[self]
 
     @property
     def provider(self) -> str:
@@ -238,125 +212,300 @@ class LlmModel(str, metaclass=LlmModelMeta):
         return self.metadata.max_output_tokens
 
 
-# Default model constant for backward compatibility
-# Uses the dynamic registry to get the default model
-DEFAULT_LLM_MODEL = LlmModel.default()
+MODEL_METADATA = {
+    # https://platform.openai.com/docs/models
+    LlmModel.O3: ModelMetadata("openai", 200000, 100000, "O3", "OpenAI", "OpenAI", 2),
+    LlmModel.O3_MINI: ModelMetadata(
+        "openai", 200000, 100000, "O3 Mini", "OpenAI", "OpenAI", 1
+    ),  # o3-mini-2025-01-31
+    LlmModel.O1: ModelMetadata(
+        "openai", 200000, 100000, "O1", "OpenAI", "OpenAI", 3
+    ),  # o1-2024-12-17
+    LlmModel.O1_MINI: ModelMetadata(
+        "openai", 128000, 65536, "O1 Mini", "OpenAI", "OpenAI", 2
+    ),  # o1-mini-2024-09-12
+    # GPT-5 models
+    LlmModel.GPT5_2: ModelMetadata(
+        "openai", 400000, 128000, "GPT-5.2", "OpenAI", "OpenAI", 3
+    ),
+    LlmModel.GPT5_1: ModelMetadata(
+        "openai", 400000, 128000, "GPT-5.1", "OpenAI", "OpenAI", 2
+    ),
+    LlmModel.GPT5: ModelMetadata(
+        "openai", 400000, 128000, "GPT-5", "OpenAI", "OpenAI", 1
+    ),
+    LlmModel.GPT5_MINI: ModelMetadata(
+        "openai", 400000, 128000, "GPT-5 Mini", "OpenAI", "OpenAI", 1
+    ),
+    LlmModel.GPT5_NANO: ModelMetadata(
+        "openai", 400000, 128000, "GPT-5 Nano", "OpenAI", "OpenAI", 1
+    ),
+    LlmModel.GPT5_CHAT: ModelMetadata(
+        "openai", 400000, 16384, "GPT-5 Chat Latest", "OpenAI", "OpenAI", 2
+    ),
+    LlmModel.GPT41: ModelMetadata(
+        "openai", 1047576, 32768, "GPT-4.1", "OpenAI", "OpenAI", 1
+    ),
+    LlmModel.GPT41_MINI: ModelMetadata(
+        "openai", 1047576, 32768, "GPT-4.1 Mini", "OpenAI", "OpenAI", 1
+    ),
+    LlmModel.GPT4O_MINI: ModelMetadata(
+        "openai", 128000, 16384, "GPT-4o Mini", "OpenAI", "OpenAI", 1
+    ),  # gpt-4o-mini-2024-07-18
+    LlmModel.GPT4O: ModelMetadata(
+        "openai", 128000, 16384, "GPT-4o", "OpenAI", "OpenAI", 2
+    ),  # gpt-4o-2024-08-06
+    LlmModel.GPT4_TURBO: ModelMetadata(
+        "openai", 128000, 4096, "GPT-4 Turbo", "OpenAI", "OpenAI", 3
+    ),  # gpt-4-turbo-2024-04-09
+    LlmModel.GPT3_5_TURBO: ModelMetadata(
+        "openai", 16385, 4096, "GPT-3.5 Turbo", "OpenAI", "OpenAI", 1
+    ),  # gpt-3.5-turbo-0125
+    # https://docs.anthropic.com/en/docs/about-claude/models
+    LlmModel.CLAUDE_4_1_OPUS: ModelMetadata(
+        "anthropic", 200000, 32000, "Claude Opus 4.1", "Anthropic", "Anthropic", 3
+    ),  # claude-opus-4-1-20250805
+    LlmModel.CLAUDE_4_OPUS: ModelMetadata(
+        "anthropic", 200000, 32000, "Claude Opus 4", "Anthropic", "Anthropic", 3
+    ),  # claude-4-opus-20250514
+    LlmModel.CLAUDE_4_SONNET: ModelMetadata(
+        "anthropic", 200000, 64000, "Claude Sonnet 4", "Anthropic", "Anthropic", 2
+    ),  # claude-4-sonnet-20250514
+    LlmModel.CLAUDE_4_6_OPUS: ModelMetadata(
+        "anthropic", 200000, 128000, "Claude Opus 4.6", "Anthropic", "Anthropic", 3
+    ),  # claude-opus-4-6
+    LlmModel.CLAUDE_4_5_OPUS: ModelMetadata(
+        "anthropic", 200000, 64000, "Claude Opus 4.5", "Anthropic", "Anthropic", 3
+    ),  # claude-opus-4-5-20251101
+    LlmModel.CLAUDE_4_5_SONNET: ModelMetadata(
+        "anthropic", 200000, 64000, "Claude Sonnet 4.5", "Anthropic", "Anthropic", 3
+    ),  # claude-sonnet-4-5-20250929
+    LlmModel.CLAUDE_4_5_HAIKU: ModelMetadata(
+        "anthropic", 200000, 64000, "Claude Haiku 4.5", "Anthropic", "Anthropic", 2
+    ),  # claude-haiku-4-5-20251001
+    LlmModel.CLAUDE_3_HAIKU: ModelMetadata(
+        "anthropic", 200000, 4096, "Claude 3 Haiku", "Anthropic", "Anthropic", 1
+    ),  # claude-3-haiku-20240307
+    # https://docs.aimlapi.com/api-overview/model-database/text-models
+    LlmModel.AIML_API_QWEN2_5_72B: ModelMetadata(
+        "aiml_api", 32000, 8000, "Qwen 2.5 72B Instruct Turbo", "AI/ML", "Qwen", 1
+    ),
+    LlmModel.AIML_API_LLAMA3_1_70B: ModelMetadata(
+        "aiml_api",
+        128000,
+        40000,
+        "Llama 3.1 Nemotron 70B Instruct",
+        "AI/ML",
+        "Nvidia",
+        1,
+    ),
+    LlmModel.AIML_API_LLAMA3_3_70B: ModelMetadata(
+        "aiml_api", 128000, None, "Llama 3.3 70B Instruct Turbo", "AI/ML", "Meta", 1
+    ),
+    LlmModel.AIML_API_META_LLAMA_3_1_70B: ModelMetadata(
+        "aiml_api", 131000, 2000, "Llama 3.1 70B Instruct Turbo", "AI/ML", "Meta", 1
+    ),
+    LlmModel.AIML_API_LLAMA_3_2_3B: ModelMetadata(
+        "aiml_api", 128000, None, "Llama 3.2 3B Instruct Turbo", "AI/ML", "Meta", 1
+    ),
+    # https://console.groq.com/docs/models
+    LlmModel.LLAMA3_3_70B: ModelMetadata(
+        "groq", 128000, 32768, "Llama 3.3 70B Versatile", "Groq", "Meta", 1
+    ),
+    LlmModel.LLAMA3_1_8B: ModelMetadata(
+        "groq", 128000, 8192, "Llama 3.1 8B Instant", "Groq", "Meta", 1
+    ),
+    # https://ollama.com/library
+    LlmModel.OLLAMA_LLAMA3_3: ModelMetadata(
+        "ollama", 8192, None, "Llama 3.3", "Ollama", "Meta", 1
+    ),
+    LlmModel.OLLAMA_LLAMA3_2: ModelMetadata(
+        "ollama", 8192, None, "Llama 3.2", "Ollama", "Meta", 1
+    ),
+    LlmModel.OLLAMA_LLAMA3_8B: ModelMetadata(
+        "ollama", 8192, None, "Llama 3", "Ollama", "Meta", 1
+    ),
+    LlmModel.OLLAMA_LLAMA3_405B: ModelMetadata(
+        "ollama", 8192, None, "Llama 3.1 405B", "Ollama", "Meta", 1
+    ),
+    LlmModel.OLLAMA_DOLPHIN: ModelMetadata(
+        "ollama", 32768, None, "Dolphin Mistral Latest", "Ollama", "Mistral AI", 1
+    ),
+    # https://openrouter.ai/models
+    LlmModel.GEMINI_2_5_PRO: ModelMetadata(
+        "open_router",
+        1050000,
+        8192,
+        "Gemini 2.5 Pro Preview 03.25",
+        "OpenRouter",
+        "Google",
+        2,
+    ),
+    LlmModel.GEMINI_3_PRO_PREVIEW: ModelMetadata(
+        "open_router", 1048576, 65535, "Gemini 3 Pro Preview", "OpenRouter", "Google", 2
+    ),
+    LlmModel.GEMINI_2_5_FLASH: ModelMetadata(
+        "open_router", 1048576, 65535, "Gemini 2.5 Flash", "OpenRouter", "Google", 1
+    ),
+    LlmModel.GEMINI_2_0_FLASH: ModelMetadata(
+        "open_router", 1048576, 8192, "Gemini 2.0 Flash 001", "OpenRouter", "Google", 1
+    ),
+    LlmModel.GEMINI_2_5_FLASH_LITE_PREVIEW: ModelMetadata(
+        "open_router",
+        1048576,
+        65535,
+        "Gemini 2.5 Flash Lite Preview 06.17",
+        "OpenRouter",
+        "Google",
+        1,
+    ),
+    LlmModel.GEMINI_2_0_FLASH_LITE: ModelMetadata(
+        "open_router",
+        1048576,
+        8192,
+        "Gemini 2.0 Flash Lite 001",
+        "OpenRouter",
+        "Google",
+        1,
+    ),
+    LlmModel.MISTRAL_NEMO: ModelMetadata(
+        "open_router", 128000, 4096, "Mistral Nemo", "OpenRouter", "Mistral AI", 1
+    ),
+    LlmModel.COHERE_COMMAND_R_08_2024: ModelMetadata(
+        "open_router", 128000, 4096, "Command R 08.2024", "OpenRouter", "Cohere", 1
+    ),
+    LlmModel.COHERE_COMMAND_R_PLUS_08_2024: ModelMetadata(
+        "open_router", 128000, 4096, "Command R Plus 08.2024", "OpenRouter", "Cohere", 2
+    ),
+    LlmModel.DEEPSEEK_CHAT: ModelMetadata(
+        "open_router", 64000, 2048, "DeepSeek Chat", "OpenRouter", "DeepSeek", 1
+    ),
+    LlmModel.DEEPSEEK_R1_0528: ModelMetadata(
+        "open_router", 163840, 163840, "DeepSeek R1 0528", "OpenRouter", "DeepSeek", 1
+    ),
+    LlmModel.PERPLEXITY_SONAR: ModelMetadata(
+        "open_router", 127000, 8000, "Sonar", "OpenRouter", "Perplexity", 1
+    ),
+    LlmModel.PERPLEXITY_SONAR_PRO: ModelMetadata(
+        "open_router", 200000, 8000, "Sonar Pro", "OpenRouter", "Perplexity", 2
+    ),
+    LlmModel.PERPLEXITY_SONAR_DEEP_RESEARCH: ModelMetadata(
+        "open_router",
+        128000,
+        16000,
+        "Sonar Deep Research",
+        "OpenRouter",
+        "Perplexity",
+        3,
+    ),
+    LlmModel.NOUSRESEARCH_HERMES_3_LLAMA_3_1_405B: ModelMetadata(
+        "open_router",
+        131000,
+        4096,
+        "Hermes 3 Llama 3.1 405B",
+        "OpenRouter",
+        "Nous Research",
+        1,
+    ),
+    LlmModel.NOUSRESEARCH_HERMES_3_LLAMA_3_1_70B: ModelMetadata(
+        "open_router",
+        12288,
+        12288,
+        "Hermes 3 Llama 3.1 70B",
+        "OpenRouter",
+        "Nous Research",
+        1,
+    ),
+    LlmModel.OPENAI_GPT_OSS_120B: ModelMetadata(
+        "open_router", 131072, 131072, "GPT-OSS 120B", "OpenRouter", "OpenAI", 1
+    ),
+    LlmModel.OPENAI_GPT_OSS_20B: ModelMetadata(
+        "open_router", 131072, 32768, "GPT-OSS 20B", "OpenRouter", "OpenAI", 1
+    ),
+    LlmModel.AMAZON_NOVA_LITE_V1: ModelMetadata(
+        "open_router", 300000, 5120, "Nova Lite V1", "OpenRouter", "Amazon", 1
+    ),
+    LlmModel.AMAZON_NOVA_MICRO_V1: ModelMetadata(
+        "open_router", 128000, 5120, "Nova Micro V1", "OpenRouter", "Amazon", 1
+    ),
+    LlmModel.AMAZON_NOVA_PRO_V1: ModelMetadata(
+        "open_router", 300000, 5120, "Nova Pro V1", "OpenRouter", "Amazon", 1
+    ),
+    LlmModel.MICROSOFT_WIZARDLM_2_8X22B: ModelMetadata(
+        "open_router", 65536, 4096, "WizardLM 2 8x22B", "OpenRouter", "Microsoft", 1
+    ),
+    LlmModel.GRYPHE_MYTHOMAX_L2_13B: ModelMetadata(
+        "open_router", 4096, 4096, "MythoMax L2 13B", "OpenRouter", "Gryphe", 1
+    ),
+    LlmModel.META_LLAMA_4_SCOUT: ModelMetadata(
+        "open_router", 131072, 131072, "Llama 4 Scout", "OpenRouter", "Meta", 1
+    ),
+    LlmModel.META_LLAMA_4_MAVERICK: ModelMetadata(
+        "open_router", 1048576, 1000000, "Llama 4 Maverick", "OpenRouter", "Meta", 1
+    ),
+    LlmModel.GROK_4: ModelMetadata(
+        "open_router", 256000, 256000, "Grok 4", "OpenRouter", "xAI", 3
+    ),
+    LlmModel.GROK_4_FAST: ModelMetadata(
+        "open_router", 2000000, 30000, "Grok 4 Fast", "OpenRouter", "xAI", 1
+    ),
+    LlmModel.GROK_4_1_FAST: ModelMetadata(
+        "open_router", 2000000, 30000, "Grok 4.1 Fast", "OpenRouter", "xAI", 1
+    ),
+    LlmModel.GROK_CODE_FAST_1: ModelMetadata(
+        "open_router", 256000, 10000, "Grok Code Fast 1", "OpenRouter", "xAI", 1
+    ),
+    LlmModel.KIMI_K2: ModelMetadata(
+        "open_router", 131000, 131000, "Kimi K2", "OpenRouter", "Moonshot AI", 1
+    ),
+    LlmModel.QWEN3_235B_A22B_THINKING: ModelMetadata(
+        "open_router",
+        262144,
+        262144,
+        "Qwen 3 235B A22B Thinking 2507",
+        "OpenRouter",
+        "Qwen",
+        1,
+    ),
+    LlmModel.QWEN3_CODER: ModelMetadata(
+        "open_router", 262144, 262144, "Qwen 3 Coder", "OpenRouter", "Qwen", 3
+    ),
+    # Llama API models
+    LlmModel.LLAMA_API_LLAMA_4_SCOUT: ModelMetadata(
+        "llama_api",
+        128000,
+        4028,
+        "Llama 4 Scout 17B 16E Instruct FP8",
+        "Llama API",
+        "Meta",
+        1,
+    ),
+    LlmModel.LLAMA_API_LLAMA4_MAVERICK: ModelMetadata(
+        "llama_api",
+        128000,
+        4028,
+        "Llama 4 Maverick 17B 128E Instruct FP8",
+        "Llama API",
+        "Meta",
+        1,
+    ),
+    LlmModel.LLAMA_API_LLAMA3_3_8B: ModelMetadata(
+        "llama_api", 128000, 4028, "Llama 3.3 8B Instruct", "Llama API", "Meta", 1
+    ),
+    LlmModel.LLAMA_API_LLAMA3_3_70B: ModelMetadata(
+        "llama_api", 128000, 4028, "Llama 3.3 70B Instruct", "Llama API", "Meta", 1
+    ),
+    # v0 by Vercel models
+    LlmModel.V0_1_5_MD: ModelMetadata("v0", 128000, 64000, "v0 1.5 MD", "V0", "V0", 1),
+    LlmModel.V0_1_5_LG: ModelMetadata("v0", 512000, 64000, "v0 1.5 LG", "V0", "V0", 1),
+    LlmModel.V0_1_0_MD: ModelMetadata("v0", 128000, 64000, "v0 1.0 MD", "V0", "V0", 1),
+}
 
+DEFAULT_LLM_MODEL = LlmModel.GPT5_2
 
-class ModelUnavailableError(ValueError):
-    """Raised when a requested LLM model cannot be resolved for use."""
-
-    pass
-
-
-@dataclass
-class ResolvedModel:
-    """Result of resolving a model for an LLM call."""
-
-    slug: str  # The actual model slug to use (may differ from requested if fallback)
-    provider: str
-    context_window: int
-    max_output_tokens: int
-    used_fallback: bool = False
-    original_slug: str | None = None  # Set if fallback was used
-
-
-async def resolve_model_for_call(llm_model: LlmModel) -> ResolvedModel:
-    """
-    Resolve a model for use in an LLM call.
-
-    Handles:
-    - Checking if the model exists in the registry
-    - Falling back to an enabled model from the same provider if disabled
-    - Refreshing the registry cache if model not found (with DB access)
-
-    Args:
-        llm_model: The requested LlmModel
-
-    Returns:
-        ResolvedModel with all necessary metadata for the call
-
-    Raises:
-        ModelUnavailableError: If model cannot be resolved (not found, disabled with no fallback)
-    """
-    from backend.data.llm_registry import (
-        get_fallback_model_for_disabled,
-        get_model_info,
-    )
-
-    model_info = get_model_info(llm_model.value)
-
-    # Case 1: Model found and disabled - try fallback
-    if model_info and not model_info.is_enabled:
-        fallback = get_fallback_model_for_disabled(llm_model.value)
-        if fallback:
-            logger.warning(
-                f"Model '{llm_model.value}' is disabled. Using fallback "
-                f"'{fallback.slug}' from same provider ({fallback.metadata.provider})."
-            )
-            return ResolvedModel(
-                slug=fallback.slug,
-                provider=fallback.metadata.provider,
-                context_window=fallback.metadata.context_window,
-                max_output_tokens=fallback.metadata.max_output_tokens or 2**15,
-                used_fallback=True,
-                original_slug=llm_model.value,
-            )
-        raise ModelUnavailableError(
-            f"Model '{llm_model.value}' is disabled and no fallback from the same "
-            f"provider is available. Enable the model or select a different one."
-        )
-
-    # Case 2: Model found and enabled - use it directly
-    if model_info:
-        return ResolvedModel(
-            slug=llm_model.value,
-            provider=model_info.metadata.provider,
-            context_window=model_info.metadata.context_window,
-            max_output_tokens=model_info.metadata.max_output_tokens or 2**15,
-        )
-
-    # Case 3: Model not in registry - try refresh if DB available
-    logger.warning(f"Model '{llm_model.value}' not found in registry cache")
-
-    from backend.data.db import is_connected
-
-    if not is_connected():
-        raise ModelUnavailableError(
-            f"Model '{llm_model.value}' not found in registry. "
-            f"The registry may need to be refreshed via the admin UI."
-        )
-
-    # Try refreshing the registry
-    try:
-        logger.info(f"Refreshing LLM registry for model '{llm_model.value}'")
-        await llm_registry.refresh_llm_registry()
-    except Exception as e:
-        raise ModelUnavailableError(
-            f"Model '{llm_model.value}' not found and registry refresh failed: {e}"
-        ) from e
-
-    # Check again after refresh
-    model_info = get_model_info(llm_model.value)
-    if not model_info:
-        raise ModelUnavailableError(
-            f"Model '{llm_model.value}' not found in registry. "
-            f"Add it via the admin UI at /admin/llms."
-        )
-
-    if not model_info.is_enabled:
-        raise ModelUnavailableError(
-            f"Model '{llm_model.value}' exists but is disabled. "
-            f"Enable it via the admin UI at /admin/llms."
-        )
-
-    logger.info(f"Model '{llm_model.value}' loaded after registry refresh")
-    return ResolvedModel(
-        slug=llm_model.value,
-        provider=model_info.metadata.provider,
-        context_window=model_info.metadata.context_window,
-        max_output_tokens=model_info.metadata.max_output_tokens or 2**15,
-    )
+for model in LlmModel:
+    if model not in MODEL_METADATA:
+        raise ValueError(f"Missing MODEL_METADATA metadata for model: {model}")
 
 
 class ToolCall(BaseModel):
@@ -382,12 +531,12 @@ class LLMResponse(BaseModel):
 
 def convert_openai_tool_fmt_to_anthropic(
     openai_tools: list[dict] | None = None,
-) -> Iterable[ToolParam] | anthropic.NotGiven:
+) -> Iterable[ToolParam] | anthropic.Omit:
     """
     Convert OpenAI tool format to Anthropic tool format.
     """
     if not openai_tools or len(openai_tools) == 0:
-        return anthropic.NOT_GIVEN
+        return anthropic.omit
 
     anthropic_tools = []
     for tool in openai_tools:
@@ -449,12 +598,7 @@ def get_parallel_tool_calls_param(
     llm_model: LlmModel, parallel_tool_calls: bool | None
 ) -> bool | openai.Omit:
     """Get the appropriate parallel_tool_calls parameter for OpenAI-compatible APIs."""
-    # Check for o-series models (o1, o1-mini, o3-mini, etc.) which don't support
-    # parallel tool calls. Handle both bare slugs ("o1-mini") and provider-prefixed
-    # slugs ("openai/o1-mini"). The pattern matches "o" followed by a digit at the
-    # start of the string or after a "/" separator.
-    is_o_series = re.search(r"(^|/)o\d", llm_model) is not None
-    if is_o_series or parallel_tool_calls is None:
+    if llm_model.startswith("o") or parallel_tool_calls is None:
         return openai.omit
     return parallel_tool_calls
 
@@ -490,22 +634,15 @@ async def llm_call(
             - prompt_tokens: The number of tokens used in the prompt.
             - completion_tokens: The number of tokens used in the completion.
     """
-    # Resolve the model - handles disabled models, fallbacks, and cache misses
-    resolved = await resolve_model_for_call(llm_model)
-
-    model_to_use = resolved.slug
-    provider = resolved.provider
-    context_window = resolved.context_window
-    model_max_output = resolved.max_output_tokens
-
-    # Create effective model for model-specific parameter resolution (e.g., o-series check)
-    effective_model = LlmModel(model_to_use)
+    provider = llm_model.metadata.provider
+    context_window = llm_model.context_window
 
     if compress_prompt_to_fit:
         result = await compress_context(
             messages=prompt,
-            target_tokens=context_window // 2,
+            target_tokens=llm_model.context_window // 2,
             client=None,  # Truncation-only, no LLM summarization
+            reserve=0,  # Caller handles response token budget separately
         )
         if result.error:
             logger.warning(
@@ -516,7 +653,7 @@ async def llm_call(
 
     # Calculate available tokens based on context window and input length
     estimated_input_tokens = estimate_token_count(prompt)
-    # model_max_output already set above
+    model_max_output = llm_model.max_output_tokens or int(2**15)
     user_max = max_tokens or model_max_output
     available_tokens = max(context_window - estimated_input_tokens, 0)
     max_tokens = max(min(available_tokens, model_max_output, user_max), 1)
@@ -527,14 +664,14 @@ async def llm_call(
         response_format = None
 
         parallel_tool_calls = get_parallel_tool_calls_param(
-            effective_model, parallel_tool_calls
+            llm_model, parallel_tool_calls
         )
 
         if force_json_output:
             response_format = {"type": "json_object"}
 
         response = await oai_client.chat.completions.create(
-            model=model_to_use,
+            model=llm_model.value,
             messages=prompt,  # type: ignore
             response_format=response_format,  # type: ignore
             max_completion_tokens=max_tokens,
@@ -581,7 +718,7 @@ async def llm_call(
         )
         try:
             resp = await client.messages.create(
-                model=model_to_use,
+                model=llm_model.value,
                 system=sysprompt,
                 messages=messages,
                 max_tokens=max_tokens,
@@ -645,7 +782,7 @@ async def llm_call(
         client = AsyncGroq(api_key=credentials.api_key.get_secret_value())
         response_format = {"type": "json_object"} if force_json_output else None
         response = await client.chat.completions.create(
-            model=model_to_use,
+            model=llm_model.value,
             messages=prompt,  # type: ignore
             response_format=response_format,  # type: ignore
             max_tokens=max_tokens,
@@ -667,7 +804,7 @@ async def llm_call(
         sys_messages = [p["content"] for p in prompt if p["role"] == "system"]
         usr_messages = [p["content"] for p in prompt if p["role"] != "system"]
         response = await client.generate(
-            model=model_to_use,
+            model=llm_model.value,
             prompt=f"{sys_messages}\n\n{usr_messages}",
             stream=False,
             options={"num_ctx": max_tokens},
@@ -689,7 +826,7 @@ async def llm_call(
         )
 
         parallel_tool_calls_param = get_parallel_tool_calls_param(
-            effective_model, parallel_tool_calls
+            llm_model, parallel_tool_calls
         )
 
         response = await client.chat.completions.create(
@@ -697,7 +834,7 @@ async def llm_call(
                 "HTTP-Referer": "https://agpt.co",
                 "X-Title": "AutoGPT",
             },
-            model=model_to_use,
+            model=llm_model.value,
             messages=prompt,  # type: ignore
             max_tokens=max_tokens,
             tools=tools_param,  # type: ignore
@@ -731,7 +868,7 @@ async def llm_call(
         )
 
         parallel_tool_calls_param = get_parallel_tool_calls_param(
-            effective_model, parallel_tool_calls
+            llm_model, parallel_tool_calls
         )
 
         response = await client.chat.completions.create(
@@ -739,7 +876,7 @@ async def llm_call(
                 "HTTP-Referer": "https://agpt.co",
                 "X-Title": "AutoGPT",
             },
-            model=model_to_use,
+            model=llm_model.value,
             messages=prompt,  # type: ignore
             max_tokens=max_tokens,
             tools=tools_param,  # type: ignore
@@ -766,7 +903,7 @@ async def llm_call(
             reasoning=reasoning,
         )
     elif provider == "aiml_api":
-        client = openai.AsyncOpenAI(
+        client = openai.OpenAI(
             base_url="https://api.aimlapi.com/v2",
             api_key=credentials.api_key.get_secret_value(),
             default_headers={
@@ -776,8 +913,8 @@ async def llm_call(
             },
         )
 
-        completion = await client.chat.completions.create(
-            model=model_to_use,
+        completion = client.chat.completions.create(
+            model=llm_model.value,
             messages=prompt,  # type: ignore
             max_tokens=max_tokens,
         )
@@ -805,11 +942,11 @@ async def llm_call(
             response_format = {"type": "json_object"}
 
         parallel_tool_calls_param = get_parallel_tool_calls_param(
-            effective_model, parallel_tool_calls
+            llm_model, parallel_tool_calls
         )
 
         response = await client.chat.completions.create(
-            model=model_to_use,
+            model=llm_model.value,
             messages=prompt,  # type: ignore
             response_format=response_format,  # type: ignore
             max_tokens=max_tokens,
@@ -860,10 +997,9 @@ class AIStructuredResponseGeneratorBlock(AIBlockBase):
         )
         model: LlmModel = SchemaField(
             title="LLM Model",
-            default_factory=LlmModel.default,
+            default=DEFAULT_LLM_MODEL,
             description="The language model to use for answering the prompt.",
             advanced=False,
-            json_schema_extra=llm_model_schema_extra(),
         )
         force_json_output: bool = SchemaField(
             title="Restrict LLM to pure JSON output",
@@ -926,7 +1062,7 @@ class AIStructuredResponseGeneratorBlock(AIBlockBase):
             input_schema=AIStructuredResponseGeneratorBlock.Input,
             output_schema=AIStructuredResponseGeneratorBlock.Output,
             test_input={
-                "model": "gpt-4o",  # Using string value - enum accepts any model slug dynamically
+                "model": DEFAULT_LLM_MODEL,
                 "credentials": TEST_CREDENTIALS_INPUT,
                 "expected_format": {
                     "key1": "value1",
@@ -1292,10 +1428,9 @@ class AITextGeneratorBlock(AIBlockBase):
         )
         model: LlmModel = SchemaField(
             title="LLM Model",
-            default_factory=LlmModel.default,
+            default=DEFAULT_LLM_MODEL,
             description="The language model to use for answering the prompt.",
             advanced=False,
-            json_schema_extra=llm_model_schema_extra(),
         )
         credentials: AICredentials = AICredentialsField()
         sys_prompt: str = SchemaField(
@@ -1389,9 +1524,8 @@ class AITextSummarizerBlock(AIBlockBase):
         )
         model: LlmModel = SchemaField(
             title="LLM Model",
-            default_factory=LlmModel.default,
+            default=DEFAULT_LLM_MODEL,
             description="The language model to use for summarizing the text.",
-            json_schema_extra=llm_model_schema_extra(),
         )
         focus: str = SchemaField(
             title="Focus",
@@ -1607,9 +1741,8 @@ class AIConversationBlock(AIBlockBase):
         )
         model: LlmModel = SchemaField(
             title="LLM Model",
-            default_factory=LlmModel.default,
+            default=DEFAULT_LLM_MODEL,
             description="The language model to use for the conversation.",
-            json_schema_extra=llm_model_schema_extra(),
         )
         credentials: AICredentials = AICredentialsField()
         max_tokens: int | None = SchemaField(
@@ -1646,7 +1779,7 @@ class AIConversationBlock(AIBlockBase):
                     },
                     {"role": "user", "content": "Where was it played?"},
                 ],
-                "model": "gpt-4o",  # Using string value - enum accepts any model slug dynamically
+                "model": DEFAULT_LLM_MODEL,
                 "credentials": TEST_CREDENTIALS_INPUT,
             },
             test_credentials=TEST_CREDENTIALS,
@@ -1709,10 +1842,9 @@ class AIListGeneratorBlock(AIBlockBase):
         )
         model: LlmModel = SchemaField(
             title="LLM Model",
-            default_factory=LlmModel.default,
+            default=DEFAULT_LLM_MODEL,
             description="The language model to use for generating the list.",
             advanced=True,
-            json_schema_extra=llm_model_schema_extra(),
         )
         credentials: AICredentials = AICredentialsField()
         max_retries: int = SchemaField(
@@ -1767,7 +1899,7 @@ class AIListGeneratorBlock(AIBlockBase):
                     "drawing explorers to uncover its mysteries. Each planet showcases the limitless possibilities of "
                     "fictional worlds."
                 ),
-                "model": "gpt-4o",  # Using string value - enum accepts any model slug dynamically
+                "model": DEFAULT_LLM_MODEL,
                 "credentials": TEST_CREDENTIALS_INPUT,
                 "max_retries": 3,
                 "force_json_output": False,

autogpt_platform/backend/backend/blocks/mcp/block.py

@@ -0,0 +1,300 @@
+"""
+MCP (Model Context Protocol) Tool Block.
+
+A single dynamic block that can connect to any MCP server, discover available tools,
+and execute them. Works like AgentExecutorBlock — the user selects a tool from a
+dropdown and the input/output schema adapts dynamically.
+"""
+
+import json
+import logging
+from typing import Any, Literal
+
+from pydantic import SecretStr
+
+from backend.blocks._base import (
+    Block,
+    BlockCategory,
+    BlockSchemaInput,
+    BlockSchemaOutput,
+    BlockType,
+)
+from backend.blocks.mcp.client import MCPClient, MCPClientError
+from backend.data.block import BlockInput, BlockOutput
+from backend.data.model import (
+    CredentialsField,
+    CredentialsMetaInput,
+    OAuth2Credentials,
+    SchemaField,
+)
+from backend.integrations.providers import ProviderName
+from backend.util.json import validate_with_jsonschema
+
+logger = logging.getLogger(__name__)
+
+TEST_CREDENTIALS = OAuth2Credentials(
+    id="test-mcp-cred",
+    provider="mcp",
+    access_token=SecretStr("mock-mcp-token"),
+    refresh_token=SecretStr("mock-refresh"),
+    scopes=[],
+    title="Mock MCP credential",
+)
+TEST_CREDENTIALS_INPUT = {
+    "provider": TEST_CREDENTIALS.provider,
+    "id": TEST_CREDENTIALS.id,
+    "type": TEST_CREDENTIALS.type,
+    "title": TEST_CREDENTIALS.title,
+}
+
+
+MCPCredentials = CredentialsMetaInput[Literal[ProviderName.MCP], Literal["oauth2"]]
+
+
+class MCPToolBlock(Block):
+    """
+    A block that connects to an MCP server, lets the user pick a tool,
+    and executes it with dynamic input/output schema.
+
+    The flow:
+    1. User provides an MCP server URL (and optional credentials)
+    2. Frontend calls the backend to get tool list from that URL
+    3. User selects a tool from a dropdown (available_tools)
+    4. The block's input schema updates to reflect the selected tool's parameters
+    5. On execution, the block calls the MCP server to run the tool
+    """
+
+    class Input(BlockSchemaInput):
+        server_url: str = SchemaField(
+            description="URL of the MCP server (Streamable HTTP endpoint)",
+            placeholder="https://mcp.example.com/mcp",
+        )
+        credentials: MCPCredentials = CredentialsField(
+            discriminator="server_url",
+            description="MCP server OAuth credentials",
+            default={},
+        )
+        selected_tool: str = SchemaField(
+            description="The MCP tool to execute",
+            placeholder="Select a tool",
+            default="",
+        )
+        tool_input_schema: dict[str, Any] = SchemaField(
+            description="JSON Schema for the selected tool's input parameters. "
+            "Populated automatically when a tool is selected.",
+            default={},
+            hidden=True,
+        )
+
+        tool_arguments: dict[str, Any] = SchemaField(
+            description="Arguments to pass to the selected MCP tool. "
+            "The fields here are defined by the tool's input schema.",
+            default={},
+        )
+
+        @classmethod
+        def get_input_schema(cls, data: BlockInput) -> dict[str, Any]:
+            """Return the tool's input schema so the builder UI renders dynamic fields."""
+            return data.get("tool_input_schema", {})
+
+        @classmethod
+        def get_input_defaults(cls, data: BlockInput) -> BlockInput:
+            """Return the current tool_arguments as defaults for the dynamic fields."""
+            return data.get("tool_arguments", {})
+
+        @classmethod
+        def get_missing_input(cls, data: BlockInput) -> set[str]:
+            """Check which required tool arguments are missing."""
+            required_fields = cls.get_input_schema(data).get("required", [])
+            tool_arguments = data.get("tool_arguments", {})
+            return set(required_fields) - set(tool_arguments)
+
+        @classmethod
+        def get_mismatch_error(cls, data: BlockInput) -> str | None:
+            """Validate tool_arguments against the tool's input schema."""
+            tool_schema = cls.get_input_schema(data)
+            if not tool_schema:
+                return None
+            tool_arguments = data.get("tool_arguments", {})
+            return validate_with_jsonschema(tool_schema, tool_arguments)
+
+    class Output(BlockSchemaOutput):
+        result: Any = SchemaField(description="The result returned by the MCP tool")
+        error: str = SchemaField(description="Error message if the tool call failed")
+
+    def __init__(self):
+        super().__init__(
+            id="a0a4b1c2-d3e4-4f56-a7b8-c9d0e1f2a3b4",
+            description="Connect to any MCP server and execute its tools. "
+            "Provide a server URL, select a tool, and pass arguments dynamically.",
+            categories={BlockCategory.DEVELOPER_TOOLS},
+            input_schema=MCPToolBlock.Input,
+            output_schema=MCPToolBlock.Output,
+            block_type=BlockType.MCP_TOOL,
+            test_credentials=TEST_CREDENTIALS,
+            test_input={
+                "server_url": "https://mcp.example.com/mcp",
+                "credentials": TEST_CREDENTIALS_INPUT,
+                "selected_tool": "get_weather",
+                "tool_input_schema": {
+                    "type": "object",
+                    "properties": {"city": {"type": "string"}},
+                    "required": ["city"],
+                },
+                "tool_arguments": {"city": "London"},
+            },
+            test_output=[
+                (
+                    "result",
+                    {"weather": "sunny", "temperature": 20},
+                ),
+            ],
+            test_mock={
+                "_call_mcp_tool": lambda *a, **kw: {
+                    "weather": "sunny",
+                    "temperature": 20,
+                },
+            },
+        )
+
+    async def _call_mcp_tool(
+        self,
+        server_url: str,
+        tool_name: str,
+        arguments: dict[str, Any],
+        auth_token: str | None = None,
+    ) -> Any:
+        """Call a tool on the MCP server. Extracted for easy mocking in tests."""
+        client = MCPClient(server_url, auth_token=auth_token)
+        await client.initialize()
+        result = await client.call_tool(tool_name, arguments)
+
+        if result.is_error:
+            error_text = ""
+            for item in result.content:
+                if item.get("type") == "text":
+                    error_text += item.get("text", "")
+            raise MCPClientError(
+                f"MCP tool '{tool_name}' returned an error: "
+                f"{error_text or 'Unknown error'}"
+            )
+
+        # Extract text content from the result
+        output_parts = []
+        for item in result.content:
+            if item.get("type") == "text":
+                text = item.get("text", "")
+                # Try to parse as JSON for structured output
+                try:
+                    output_parts.append(json.loads(text))
+                except (json.JSONDecodeError, ValueError):
+                    output_parts.append(text)
+            elif item.get("type") == "image":
+                output_parts.append(
+                    {
+                        "type": "image",
+                        "data": item.get("data"),
+                        "mimeType": item.get("mimeType"),
+                    }
+                )
+            elif item.get("type") == "resource":
+                output_parts.append(item.get("resource", {}))
+
+        # If single result, unwrap
+        if len(output_parts) == 1:
+            return output_parts[0]
+        return output_parts if output_parts else None
+
+    @staticmethod
+    async def _auto_lookup_credential(
+        user_id: str, server_url: str
+    ) -> "OAuth2Credentials | None":
+        """Auto-lookup stored MCP credential for a server URL.
+
+        This is a fallback for nodes that don't have ``credentials`` explicitly
+        set (e.g. nodes created before the credential field was wired up).
+        """
+        from backend.integrations.creds_manager import IntegrationCredentialsManager
+        from backend.integrations.providers import ProviderName
+
+        try:
+            mgr = IntegrationCredentialsManager()
+            mcp_creds = await mgr.store.get_creds_by_provider(
+                user_id, ProviderName.MCP.value
+            )
+            best: OAuth2Credentials | None = None
+            for cred in mcp_creds:
+                if (
+                    isinstance(cred, OAuth2Credentials)
+                    and (cred.metadata or {}).get("mcp_server_url") == server_url
+                ):
+                    if best is None or (
+                        (cred.access_token_expires_at or 0)
+                        > (best.access_token_expires_at or 0)
+                    ):
+                        best = cred
+            if best:
+                best = await mgr.refresh_if_needed(user_id, best)
+                logger.info(
+                    "Auto-resolved MCP credential %s for %s", best.id, server_url
+                )
+            return best
+        except Exception:
+            logger.warning("Auto-lookup MCP credential failed", exc_info=True)
+            return None
+
+    async def run(
+        self,
+        input_data: Input,
+        *,
+        user_id: str,
+        credentials: OAuth2Credentials | None = None,
+        **kwargs,
+    ) -> BlockOutput:
+        if not input_data.server_url:
+            yield "error", "MCP server URL is required"
+            return
+
+        if not input_data.selected_tool:
+            yield "error", "No tool selected. Please select a tool from the dropdown."
+            return
+
+        # Validate required tool arguments before calling the server.
+        # The executor-level validation is bypassed for MCP blocks because
+        # get_input_defaults() flattens tool_arguments, stripping tool_input_schema
+        # from the validation context.
+        required = set(input_data.tool_input_schema.get("required", []))
+        if required:
+            missing = required - set(input_data.tool_arguments.keys())
+            if missing:
+                yield "error", (
+                    f"Missing required argument(s): {', '.join(sorted(missing))}. "
+                    f"Please fill in all required fields marked with * in the block form."
+                )
+                return
+
+        # If no credentials were injected by the executor (e.g. legacy nodes
+        # that don't have the credentials field set), try to auto-lookup
+        # the stored MCP credential for this server URL.
+        if credentials is None:
+            credentials = await self._auto_lookup_credential(
+                user_id, input_data.server_url
+            )
+
+        auth_token = (
+            credentials.access_token.get_secret_value() if credentials else None
+        )
+
+        try:
+            result = await self._call_mcp_tool(
+                server_url=input_data.server_url,
+                tool_name=input_data.selected_tool,
+                arguments=input_data.tool_arguments,
+                auth_token=auth_token,
+            )
+            yield "result", result
+        except MCPClientError as e:
+            yield "error", str(e)
+        except Exception as e:
+            logger.exception(f"MCP tool call failed: {e}")
+            yield "error", f"MCP tool call failed: {str(e)}"

autogpt_platform/backend/backend/blocks/mcp/client.py

@@ -0,0 +1,323 @@
+"""
+MCP (Model Context Protocol) HTTP client.
+
+Implements the MCP Streamable HTTP transport for listing tools and calling tools
+on remote MCP servers. Uses JSON-RPC 2.0 over HTTP POST.
+
+Handles both JSON and SSE (text/event-stream) response formats per the MCP spec.
+
+Reference: https://modelcontextprotocol.io/specification/2025-03-26/basic/transports
+"""
+
+import json
+import logging
+from dataclasses import dataclass, field
+from typing import Any
+
+from backend.util.request import Requests
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class MCPTool:
+    """Represents an MCP tool discovered from a server."""
+
+    name: str
+    description: str
+    input_schema: dict[str, Any]
+
+
+@dataclass
+class MCPCallResult:
+    """Result from calling an MCP tool."""
+
+    content: list[dict[str, Any]] = field(default_factory=list)
+    is_error: bool = False
+
+
+class MCPClientError(Exception):
+    """Raised when an MCP protocol error occurs."""
+
+    pass
+
+
+class MCPClient:
+    """
+    Async HTTP client for the MCP Streamable HTTP transport.
+
+    Communicates with MCP servers using JSON-RPC 2.0 over HTTP POST.
+    Supports optional Bearer token authentication.
+    """
+
+    def __init__(
+        self,
+        server_url: str,
+        auth_token: str | None = None,
+    ):
+        self.server_url = server_url.rstrip("/")
+        self.auth_token = auth_token
+        self._request_id = 0
+        self._session_id: str | None = None
+
+    def _next_id(self) -> int:
+        self._request_id += 1
+        return self._request_id
+
+    def _build_headers(self) -> dict[str, str]:
+        headers = {
+            "Content-Type": "application/json",
+            "Accept": "application/json, text/event-stream",
+        }
+        if self.auth_token:
+            headers["Authorization"] = f"Bearer {self.auth_token}"
+        if self._session_id:
+            headers["Mcp-Session-Id"] = self._session_id
+        return headers
+
+    def _build_jsonrpc_request(
+        self, method: str, params: dict[str, Any] | None = None
+    ) -> dict[str, Any]:
+        req: dict[str, Any] = {
+            "jsonrpc": "2.0",
+            "method": method,
+            "id": self._next_id(),
+        }
+        if params is not None:
+            req["params"] = params
+        return req
+
+    @staticmethod
+    def _parse_sse_response(text: str) -> dict[str, Any]:
+        """Parse an SSE (text/event-stream) response body into JSON-RPC data.
+
+        MCP servers may return responses as SSE with format:
+            event: message
+            data: {"jsonrpc":"2.0","result":{...},"id":1}
+
+        We extract the last `data:` line that contains a JSON-RPC response
+        (i.e. has an "id" field), which is the reply to our request.
+        """
+        last_data: dict[str, Any] | None = None
+        for line in text.splitlines():
+            stripped = line.strip()
+            if stripped.startswith("data:"):
+                payload = stripped[len("data:") :].strip()
+                if not payload:
+                    continue
+                try:
+                    parsed = json.loads(payload)
+                    # Only keep JSON-RPC responses (have "id"), skip notifications
+                    if isinstance(parsed, dict) and "id" in parsed:
+                        last_data = parsed
+                except (json.JSONDecodeError, ValueError):
+                    continue
+        if last_data is None:
+            raise MCPClientError("No JSON-RPC response found in SSE stream")
+        return last_data
+
+    async def _send_request(
+        self, method: str, params: dict[str, Any] | None = None
+    ) -> Any:
+        """Send a JSON-RPC request to the MCP server and return the result.
+
+        Handles both ``application/json`` and ``text/event-stream`` responses
+        as required by the MCP Streamable HTTP transport specification.
+        """
+        payload = self._build_jsonrpc_request(method, params)
+        headers = self._build_headers()
+
+        requests = Requests(
+            raise_for_status=True,
+            extra_headers=headers,
+        )
+        response = await requests.post(self.server_url, json=payload)
+
+        # Capture session ID from response (MCP Streamable HTTP transport)
+        session_id = response.headers.get("Mcp-Session-Id")
+        if session_id:
+            self._session_id = session_id
+
+        content_type = response.headers.get("content-type", "")
+        if "text/event-stream" in content_type:
+            body = self._parse_sse_response(response.text())
+        else:
+            try:
+                body = response.json()
+            except Exception as e:
+                raise MCPClientError(
+                    f"MCP server returned non-JSON response: {e}"
+                ) from e
+
+        if not isinstance(body, dict):
+            raise MCPClientError(
+                f"MCP server returned unexpected JSON type: {type(body).__name__}"
+            )
+
+        # Handle JSON-RPC error
+        if "error" in body:
+            error = body["error"]
+            if isinstance(error, dict):
+                raise MCPClientError(
+                    f"MCP server error [{error.get('code', '?')}]: "
+                    f"{error.get('message', 'Unknown error')}"
+                )
+            raise MCPClientError(f"MCP server error: {error}")
+
+        return body.get("result")
+
+    async def _send_notification(self, method: str) -> None:
+        """Send a JSON-RPC notification (no id, no response expected)."""
+        headers = self._build_headers()
+        notification = {"jsonrpc": "2.0", "method": method}
+        requests = Requests(
+            raise_for_status=False,
+            extra_headers=headers,
+        )
+        await requests.post(self.server_url, json=notification)
+
+    async def discover_auth(self) -> dict[str, Any] | None:
+        """Probe the MCP server's OAuth metadata (RFC 9728 / MCP spec).
+
+        Returns ``None`` if the server doesn't require auth, otherwise returns
+        a dict with:
+          - ``authorization_servers``: list of authorization server URLs
+          - ``resource``: the resource indicator URL (usually the MCP endpoint)
+          - ``scopes_supported``: optional list of supported scopes
+
+        The caller can then fetch the authorization server metadata to get
+        ``authorization_endpoint``, ``token_endpoint``, etc.
+        """
+        from urllib.parse import urlparse
+
+        parsed = urlparse(self.server_url)
+        base = f"{parsed.scheme}://{parsed.netloc}"
+
+        # Build candidates for protected-resource metadata (per RFC 9728)
+        path = parsed.path.rstrip("/")
+        candidates = []
+        if path and path != "/":
+            candidates.append(f"{base}/.well-known/oauth-protected-resource{path}")
+        candidates.append(f"{base}/.well-known/oauth-protected-resource")
+
+        requests = Requests(
+            raise_for_status=False,
+        )
+        for url in candidates:
+            try:
+                resp = await requests.get(url)
+                if resp.status == 200:
+                    data = resp.json()
+                    if isinstance(data, dict) and "authorization_servers" in data:
+                        return data
+            except Exception:
+                continue
+
+        return None
+
+    async def discover_auth_server_metadata(
+        self, auth_server_url: str
+    ) -> dict[str, Any] | None:
+        """Fetch the OAuth Authorization Server Metadata (RFC 8414).
+
+        Given an authorization server URL, returns a dict with:
+          - ``authorization_endpoint``
+          - ``token_endpoint``
+          - ``registration_endpoint`` (for dynamic client registration)
+          - ``scopes_supported``
+          - ``code_challenge_methods_supported``
+          - etc.
+        """
+        from urllib.parse import urlparse
+
+        parsed = urlparse(auth_server_url)
+        base = f"{parsed.scheme}://{parsed.netloc}"
+        path = parsed.path.rstrip("/")
+
+        # Try standard metadata endpoints (RFC 8414 and OpenID Connect)
+        candidates = []
+        if path and path != "/":
+            candidates.append(f"{base}/.well-known/oauth-authorization-server{path}")
+        candidates.append(f"{base}/.well-known/oauth-authorization-server")
+        candidates.append(f"{base}/.well-known/openid-configuration")
+
+        requests = Requests(
+            raise_for_status=False,
+        )
+        for url in candidates:
+            try:
+                resp = await requests.get(url)
+                if resp.status == 200:
+                    data = resp.json()
+                    if isinstance(data, dict) and "authorization_endpoint" in data:
+                        return data
+            except Exception:
+                continue
+
+        return None
+
+    async def initialize(self) -> dict[str, Any]:
+        """
+        Send the MCP initialize request.
+
+        This is required by the MCP protocol before any other requests.
+        Returns the server's capabilities.
+        """
+        result = await self._send_request(
+            "initialize",
+            {
+                "protocolVersion": "2025-03-26",
+                "capabilities": {},
+                "clientInfo": {"name": "AutoGPT-Platform", "version": "1.0.0"},
+            },
+        )
+        # Send initialized notification (no response expected)
+        await self._send_notification("notifications/initialized")
+
+        return result or {}
+
+    async def list_tools(self) -> list[MCPTool]:
+        """
+        Discover available tools from the MCP server.
+
+        Returns a list of MCPTool objects with name, description, and input schema.
+        """
+        result = await self._send_request("tools/list")
+        if not result or "tools" not in result:
+            return []
+
+        tools = []
+        for tool_data in result["tools"]:
+            tools.append(
+                MCPTool(
+                    name=tool_data.get("name", ""),
+                    description=tool_data.get("description", ""),
+                    input_schema=tool_data.get("inputSchema", {}),
+                )
+            )
+        return tools
+
+    async def call_tool(
+        self, tool_name: str, arguments: dict[str, Any]
+    ) -> MCPCallResult:
+        """
+        Call a tool on the MCP server.
+
+        Args:
+            tool_name: The name of the tool to call.
+            arguments: The arguments to pass to the tool.
+
+        Returns:
+            MCPCallResult with the tool's response content.
+        """
+        result = await self._send_request(
+            "tools/call",
+            {"name": tool_name, "arguments": arguments},
+        )
+        if not result:
+            return MCPCallResult(is_error=True)
+
+        return MCPCallResult(
+            content=result.get("content", []),
+            is_error=result.get("isError", False),
+        )

autogpt_platform/backend/backend/blocks/mcp/oauth.py

@@ -0,0 +1,204 @@
+"""
+MCP OAuth handler for MCP servers that use OAuth 2.1 authorization.
+
+Unlike other OAuth handlers (GitHub, Google, etc.) where endpoints are fixed,
+MCP servers have dynamic endpoints discovered via RFC 9728 / RFC 8414 metadata.
+This handler accepts those endpoints at construction time.
+"""
+
+import logging
+import time
+import urllib.parse
+from typing import ClassVar, Optional
+
+from pydantic import SecretStr
+
+from backend.data.model import OAuth2Credentials
+from backend.integrations.oauth.base import BaseOAuthHandler
+from backend.integrations.providers import ProviderName
+from backend.util.request import Requests
+
+logger = logging.getLogger(__name__)
+
+
+class MCPOAuthHandler(BaseOAuthHandler):
+    """
+    OAuth handler for MCP servers with dynamically-discovered endpoints.
+
+    Construction requires the authorization and token endpoint URLs,
+    which are obtained via MCP OAuth metadata discovery
+    (``MCPClient.discover_auth`` + ``discover_auth_server_metadata``).
+    """
+
+    PROVIDER_NAME: ClassVar[ProviderName | str] = ProviderName.MCP
+    DEFAULT_SCOPES: ClassVar[list[str]] = []
+
+    def __init__(
+        self,
+        client_id: str,
+        client_secret: str,
+        redirect_uri: str,
+        *,
+        authorize_url: str,
+        token_url: str,
+        revoke_url: str | None = None,
+        resource_url: str | None = None,
+    ):
+        self.client_id = client_id
+        self.client_secret = client_secret
+        self.redirect_uri = redirect_uri
+        self.authorize_url = authorize_url
+        self.token_url = token_url
+        self.revoke_url = revoke_url
+        self.resource_url = resource_url
+
+    def get_login_url(
+        self,
+        scopes: list[str],
+        state: str,
+        code_challenge: Optional[str],
+    ) -> str:
+        scopes = self.handle_default_scopes(scopes)
+
+        params: dict[str, str] = {
+            "response_type": "code",
+            "client_id": self.client_id,
+            "redirect_uri": self.redirect_uri,
+            "state": state,
+        }
+        if scopes:
+            params["scope"] = " ".join(scopes)
+        # PKCE (S256) — included when the caller provides a code_challenge
+        if code_challenge:
+            params["code_challenge"] = code_challenge
+            params["code_challenge_method"] = "S256"
+        # MCP spec requires resource indicator (RFC 8707)
+        if self.resource_url:
+            params["resource"] = self.resource_url
+
+        return f"{self.authorize_url}?{urllib.parse.urlencode(params)}"
+
+    async def exchange_code_for_tokens(
+        self,
+        code: str,
+        scopes: list[str],
+        code_verifier: Optional[str],
+    ) -> OAuth2Credentials:
+        data: dict[str, str] = {
+            "grant_type": "authorization_code",
+            "code": code,
+            "redirect_uri": self.redirect_uri,
+            "client_id": self.client_id,
+        }
+        if self.client_secret:
+            data["client_secret"] = self.client_secret
+        if code_verifier:
+            data["code_verifier"] = code_verifier
+        if self.resource_url:
+            data["resource"] = self.resource_url
+
+        response = await Requests(raise_for_status=True).post(
+            self.token_url,
+            data=data,
+            headers={"Content-Type": "application/x-www-form-urlencoded"},
+        )
+        tokens = response.json()
+
+        if "error" in tokens:
+            raise RuntimeError(
+                f"Token exchange failed: {tokens.get('error_description', tokens['error'])}"
+            )
+
+        if "access_token" not in tokens:
+            raise RuntimeError("OAuth token response missing 'access_token' field")
+
+        now = int(time.time())
+        expires_in = tokens.get("expires_in")
+
+        return OAuth2Credentials(
+            provider=self.PROVIDER_NAME,
+            title=None,
+            access_token=SecretStr(tokens["access_token"]),
+            refresh_token=(
+                SecretStr(tokens["refresh_token"])
+                if tokens.get("refresh_token")
+                else None
+            ),
+            access_token_expires_at=now + expires_in if expires_in else None,
+            refresh_token_expires_at=None,
+            scopes=scopes,
+            metadata={
+                "mcp_token_url": self.token_url,
+                "mcp_resource_url": self.resource_url,
+            },
+        )
+
+    async def _refresh_tokens(
+        self, credentials: OAuth2Credentials
+    ) -> OAuth2Credentials:
+        if not credentials.refresh_token:
+            raise ValueError("No refresh token available for MCP OAuth credentials")
+
+        data: dict[str, str] = {
+            "grant_type": "refresh_token",
+            "refresh_token": credentials.refresh_token.get_secret_value(),
+            "client_id": self.client_id,
+        }
+        if self.client_secret:
+            data["client_secret"] = self.client_secret
+        if self.resource_url:
+            data["resource"] = self.resource_url
+
+        response = await Requests(raise_for_status=True).post(
+            self.token_url,
+            data=data,
+            headers={"Content-Type": "application/x-www-form-urlencoded"},
+        )
+        tokens = response.json()
+
+        if "error" in tokens:
+            raise RuntimeError(
+                f"Token refresh failed: {tokens.get('error_description', tokens['error'])}"
+            )
+
+        if "access_token" not in tokens:
+            raise RuntimeError("OAuth refresh response missing 'access_token' field")
+
+        now = int(time.time())
+        expires_in = tokens.get("expires_in")
+
+        return OAuth2Credentials(
+            id=credentials.id,
+            provider=self.PROVIDER_NAME,
+            title=credentials.title,
+            access_token=SecretStr(tokens["access_token"]),
+            refresh_token=(
+                SecretStr(tokens["refresh_token"])
+                if tokens.get("refresh_token")
+                else credentials.refresh_token
+            ),
+            access_token_expires_at=now + expires_in if expires_in else None,
+            refresh_token_expires_at=credentials.refresh_token_expires_at,
+            scopes=credentials.scopes,
+            metadata=credentials.metadata,
+        )
+
+    async def revoke_tokens(self, credentials: OAuth2Credentials) -> bool:
+        if not self.revoke_url:
+            return False
+
+        try:
+            data = {
+                "token": credentials.access_token.get_secret_value(),
+                "token_type_hint": "access_token",
+                "client_id": self.client_id,
+            }
+            await Requests().post(
+                self.revoke_url,
+                data=data,
+                headers={"Content-Type": "application/x-www-form-urlencoded"},
+            )
+            return True
+        except Exception:
+            logger.warning("Failed to revoke MCP OAuth tokens", exc_info=True)
+            return False

autogpt_platform/backend/backend/blocks/mcp/test_e2e.py

@@ -0,0 +1,109 @@
+"""
+End-to-end tests against a real public MCP server.
+
+These tests hit the OpenAI docs MCP server (https://developers.openai.com/mcp)
+which is publicly accessible without authentication and returns SSE responses.
+
+Mark: These are tagged with ``@pytest.mark.e2e`` so they can be run/skipped
+independently of the rest of the test suite (they require network access).
+"""
+
+import json
+import os
+
+import pytest
+
+from backend.blocks.mcp.client import MCPClient
+
+# Public MCP server that requires no authentication
+OPENAI_DOCS_MCP_URL = "https://developers.openai.com/mcp"
+
+# Skip all tests in this module unless RUN_E2E env var is set
+pytestmark = pytest.mark.skipif(
+    not os.environ.get("RUN_E2E"), reason="set RUN_E2E=1 to run e2e tests"
+)
+
+
+class TestRealMCPServer:
+    """Tests against the live OpenAI docs MCP server."""
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_initialize(self):
+        """Verify we can complete the MCP handshake with a real server."""
+        client = MCPClient(OPENAI_DOCS_MCP_URL)
+        result = await client.initialize()
+
+        assert result["protocolVersion"] == "2025-03-26"
+        assert "serverInfo" in result
+        assert result["serverInfo"]["name"] == "openai-docs-mcp"
+        assert "tools" in result.get("capabilities", {})
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_list_tools(self):
+        """Verify we can discover tools from a real MCP server."""
+        client = MCPClient(OPENAI_DOCS_MCP_URL)
+        await client.initialize()
+        tools = await client.list_tools()
+
+        assert len(tools) >= 3  # server has at least 5 tools as of writing
+
+        tool_names = {t.name for t in tools}
+        # These tools are documented and should be stable
+        assert "search_openai_docs" in tool_names
+        assert "list_openai_docs" in tool_names
+        assert "fetch_openai_doc" in tool_names
+
+        # Verify schema structure
+        search_tool = next(t for t in tools if t.name == "search_openai_docs")
+        assert "query" in search_tool.input_schema.get("properties", {})
+        assert "query" in search_tool.input_schema.get("required", [])
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_tool_list_api_endpoints(self):
+        """Call the list_api_endpoints tool and verify we get real data."""
+        client = MCPClient(OPENAI_DOCS_MCP_URL)
+        await client.initialize()
+        result = await client.call_tool("list_api_endpoints", {})
+
+        assert not result.is_error
+        assert len(result.content) >= 1
+        assert result.content[0]["type"] == "text"
+
+        data = json.loads(result.content[0]["text"])
+        assert "paths" in data or "urls" in data
+        # The OpenAI API should have many endpoints
+        total = data.get("total", len(data.get("paths", [])))
+        assert total > 50
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_tool_search(self):
+        """Search for docs and verify we get results."""
+        client = MCPClient(OPENAI_DOCS_MCP_URL)
+        await client.initialize()
+        result = await client.call_tool(
+            "search_openai_docs", {"query": "chat completions", "limit": 3}
+        )
+
+        assert not result.is_error
+        assert len(result.content) >= 1
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_sse_response_handling(self):
+        """Verify the client correctly handles SSE responses from a real server.
+
+        This is the key test — our local test server returns JSON,
+        but real MCP servers typically return SSE. This proves the
+        SSE parsing works end-to-end.
+        """
+        client = MCPClient(OPENAI_DOCS_MCP_URL)
+        # initialize() internally calls _send_request which must parse SSE
+        result = await client.initialize()
+
+        # If we got here without error, SSE parsing works
+        assert isinstance(result, dict)
+        assert "protocolVersion" in result
+
+        # Also verify list_tools works (another SSE response)
+        tools = await client.list_tools()
+        assert len(tools) > 0
+        assert all(hasattr(t, "name") for t in tools)

autogpt_platform/backend/backend/blocks/mcp/test_integration.py

@@ -0,0 +1,389 @@
+"""
+Integration tests for MCP client and MCPToolBlock against a real HTTP server.
+
+These tests spin up a local MCP test server and run the full client/block flow
+against it — no mocking, real HTTP requests.
+"""
+
+import asyncio
+import json
+import threading
+from unittest.mock import patch
+
+import pytest
+from aiohttp import web
+from pydantic import SecretStr
+
+from backend.blocks.mcp.block import MCPToolBlock
+from backend.blocks.mcp.client import MCPClient
+from backend.blocks.mcp.test_server import create_test_mcp_app
+from backend.data.model import OAuth2Credentials
+
+MOCK_USER_ID = "test-user-integration"
+
+
+class _MCPTestServer:
+    """
+    Run an MCP test server in a background thread with its own event loop.
+    This avoids event loop conflicts with pytest-asyncio.
+    """
+
+    def __init__(self, auth_token: str | None = None):
+        self.auth_token = auth_token
+        self.url: str = ""
+        self._runner: web.AppRunner | None = None
+        self._loop: asyncio.AbstractEventLoop | None = None
+        self._thread: threading.Thread | None = None
+        self._started = threading.Event()
+
+    def _run(self):
+        self._loop = asyncio.new_event_loop()
+        asyncio.set_event_loop(self._loop)
+        self._loop.run_until_complete(self._start())
+        self._started.set()
+        self._loop.run_forever()
+
+    async def _start(self):
+        app = create_test_mcp_app(auth_token=self.auth_token)
+        self._runner = web.AppRunner(app)
+        await self._runner.setup()
+        site = web.TCPSite(self._runner, "127.0.0.1", 0)
+        await site.start()
+        port = site._server.sockets[0].getsockname()[1]  # type: ignore[union-attr]
+        self.url = f"http://127.0.0.1:{port}/mcp"
+
+    def start(self):
+        self._thread = threading.Thread(target=self._run, daemon=True)
+        self._thread.start()
+        if not self._started.wait(timeout=5):
+            raise RuntimeError("MCP test server failed to start within 5 seconds")
+        return self
+
+    def stop(self):
+        if self._loop and self._runner:
+            asyncio.run_coroutine_threadsafe(self._runner.cleanup(), self._loop).result(
+                timeout=5
+            )
+            self._loop.call_soon_threadsafe(self._loop.stop)
+        if self._thread:
+            self._thread.join(timeout=5)
+
+
+@pytest.fixture(scope="module")
+def mcp_server():
+    """Start a local MCP test server in a background thread."""
+    server = _MCPTestServer()
+    server.start()
+    yield server.url
+    server.stop()
+
+
+@pytest.fixture(scope="module")
+def mcp_server_with_auth():
+    """Start a local MCP test server with auth in a background thread."""
+    server = _MCPTestServer(auth_token="test-secret-token")
+    server.start()
+    yield server.url, "test-secret-token"
+    server.stop()
+
+
+@pytest.fixture(autouse=True)
+def _allow_localhost():
+    """
+    Allow 127.0.0.1 through SSRF protection for integration tests.
+
+    The Requests class blocks private IPs by default. We patch the Requests
+    constructor to always include 127.0.0.1 as a trusted origin so the local
+    test server is reachable.
+    """
+    from backend.util.request import Requests
+
+    original_init = Requests.__init__
+
+    def patched_init(self, *args, **kwargs):
+        trusted = list(kwargs.get("trusted_origins") or [])
+        trusted.append("http://127.0.0.1")
+        kwargs["trusted_origins"] = trusted
+        original_init(self, *args, **kwargs)
+
+    with patch.object(Requests, "__init__", patched_init):
+        yield
+
+
+def _make_client(url: str, auth_token: str | None = None) -> MCPClient:
+    """Create an MCPClient for integration tests."""
+    return MCPClient(url, auth_token=auth_token)
+
+
+# ── MCPClient integration tests ──────────────────────────────────────
+
+
+class TestMCPClientIntegration:
+    """Test MCPClient against a real local MCP server."""
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_initialize(self, mcp_server):
+        client = _make_client(mcp_server)
+        result = await client.initialize()
+
+        assert result["protocolVersion"] == "2025-03-26"
+        assert result["serverInfo"]["name"] == "test-mcp-server"
+        assert "tools" in result["capabilities"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_list_tools(self, mcp_server):
+        client = _make_client(mcp_server)
+        await client.initialize()
+        tools = await client.list_tools()
+
+        assert len(tools) == 3
+
+        tool_names = {t.name for t in tools}
+        assert tool_names == {"get_weather", "add_numbers", "echo"}
+
+        # Check get_weather schema
+        weather = next(t for t in tools if t.name == "get_weather")
+        assert weather.description == "Get current weather for a city"
+        assert "city" in weather.input_schema["properties"]
+        assert weather.input_schema["required"] == ["city"]
+
+        # Check add_numbers schema
+        add = next(t for t in tools if t.name == "add_numbers")
+        assert "a" in add.input_schema["properties"]
+        assert "b" in add.input_schema["properties"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_tool_get_weather(self, mcp_server):
+        client = _make_client(mcp_server)
+        await client.initialize()
+        result = await client.call_tool("get_weather", {"city": "London"})
+
+        assert not result.is_error
+        assert len(result.content) == 1
+        assert result.content[0]["type"] == "text"
+
+        data = json.loads(result.content[0]["text"])
+        assert data["city"] == "London"
+        assert data["temperature"] == 22
+        assert data["condition"] == "sunny"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_tool_add_numbers(self, mcp_server):
+        client = _make_client(mcp_server)
+        await client.initialize()
+        result = await client.call_tool("add_numbers", {"a": 3, "b": 7})
+
+        assert not result.is_error
+        data = json.loads(result.content[0]["text"])
+        assert data["result"] == 10
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_tool_echo(self, mcp_server):
+        client = _make_client(mcp_server)
+        await client.initialize()
+        result = await client.call_tool("echo", {"message": "Hello MCP!"})
+
+        assert not result.is_error
+        assert result.content[0]["text"] == "Hello MCP!"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_unknown_tool(self, mcp_server):
+        client = _make_client(mcp_server)
+        await client.initialize()
+        result = await client.call_tool("nonexistent_tool", {})
+
+        assert result.is_error
+        assert "Unknown tool" in result.content[0]["text"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_auth_success(self, mcp_server_with_auth):
+        url, token = mcp_server_with_auth
+        client = _make_client(url, auth_token=token)
+        result = await client.initialize()
+
+        assert result["protocolVersion"] == "2025-03-26"
+
+        tools = await client.list_tools()
+        assert len(tools) == 3
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_auth_failure(self, mcp_server_with_auth):
+        url, _ = mcp_server_with_auth
+        client = _make_client(url, auth_token="wrong-token")
+
+        with pytest.raises(Exception):
+            await client.initialize()
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_auth_missing(self, mcp_server_with_auth):
+        url, _ = mcp_server_with_auth
+        client = _make_client(url)
+
+        with pytest.raises(Exception):
+            await client.initialize()
+
+
+# ── MCPToolBlock integration tests ───────────────────────────────────
+
+
+class TestMCPToolBlockIntegration:
+    """Test MCPToolBlock end-to-end against a real local MCP server."""
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_full_flow_get_weather(self, mcp_server):
+        """Full flow: discover tools, select one, execute it."""
+        # Step 1: Discover tools (simulating what the frontend/API would do)
+        client = _make_client(mcp_server)
+        await client.initialize()
+        tools = await client.list_tools()
+        assert len(tools) == 3
+
+        # Step 2: User selects "get_weather" and we get its schema
+        weather_tool = next(t for t in tools if t.name == "get_weather")
+
+        # Step 3: Execute the block — no credentials (public server)
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url=mcp_server,
+            selected_tool="get_weather",
+            tool_input_schema=weather_tool.input_schema,
+            tool_arguments={"city": "Paris"},
+        )
+
+        outputs = []
+        async for name, data in block.run(input_data, user_id=MOCK_USER_ID):
+            outputs.append((name, data))
+
+        assert len(outputs) == 1
+        assert outputs[0][0] == "result"
+        result = outputs[0][1]
+        assert result["city"] == "Paris"
+        assert result["temperature"] == 22
+        assert result["condition"] == "sunny"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_full_flow_add_numbers(self, mcp_server):
+        """Full flow for add_numbers tool."""
+        client = _make_client(mcp_server)
+        await client.initialize()
+        tools = await client.list_tools()
+        add_tool = next(t for t in tools if t.name == "add_numbers")
+
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url=mcp_server,
+            selected_tool="add_numbers",
+            tool_input_schema=add_tool.input_schema,
+            tool_arguments={"a": 42, "b": 58},
+        )
+
+        outputs = []
+        async for name, data in block.run(input_data, user_id=MOCK_USER_ID):
+            outputs.append((name, data))
+
+        assert len(outputs) == 1
+        assert outputs[0][0] == "result"
+        assert outputs[0][1]["result"] == 100
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_full_flow_echo_plain_text(self, mcp_server):
+        """Verify plain text (non-JSON) responses work."""
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url=mcp_server,
+            selected_tool="echo",
+            tool_input_schema={
+                "type": "object",
+                "properties": {"message": {"type": "string"}},
+                "required": ["message"],
+            },
+            tool_arguments={"message": "Hello from AutoGPT!"},
+        )
+
+        outputs = []
+        async for name, data in block.run(input_data, user_id=MOCK_USER_ID):
+            outputs.append((name, data))
+
+        assert len(outputs) == 1
+        assert outputs[0][0] == "result"
+        assert outputs[0][1] == "Hello from AutoGPT!"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_full_flow_unknown_tool_yields_error(self, mcp_server):
+        """Calling an unknown tool should yield an error output."""
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url=mcp_server,
+            selected_tool="nonexistent_tool",
+            tool_arguments={},
+        )
+
+        outputs = []
+        async for name, data in block.run(input_data, user_id=MOCK_USER_ID):
+            outputs.append((name, data))
+
+        assert len(outputs) == 1
+        assert outputs[0][0] == "error"
+        assert "returned an error" in outputs[0][1]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_full_flow_with_auth(self, mcp_server_with_auth):
+        """Full flow with authentication via credentials kwarg."""
+        url, token = mcp_server_with_auth
+
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url=url,
+            selected_tool="echo",
+            tool_input_schema={
+                "type": "object",
+                "properties": {"message": {"type": "string"}},
+                "required": ["message"],
+            },
+            tool_arguments={"message": "Authenticated!"},
+        )
+
+        # Pass credentials via the standard kwarg (as the executor would)
+        test_creds = OAuth2Credentials(
+            id="test-cred",
+            provider="mcp",
+            access_token=SecretStr(token),
+            refresh_token=SecretStr(""),
+            scopes=[],
+            title="Test MCP credential",
+        )
+
+        outputs = []
+        async for name, data in block.run(
+            input_data, user_id=MOCK_USER_ID, credentials=test_creds
+        ):
+            outputs.append((name, data))
+
+        assert len(outputs) == 1
+        assert outputs[0][0] == "result"
+        assert outputs[0][1] == "Authenticated!"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_no_credentials_runs_without_auth(self, mcp_server):
+        """Block runs without auth when no credentials are provided."""
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url=mcp_server,
+            selected_tool="echo",
+            tool_input_schema={
+                "type": "object",
+                "properties": {"message": {"type": "string"}},
+                "required": ["message"],
+            },
+            tool_arguments={"message": "No auth needed"},
+        )
+
+        outputs = []
+        async for name, data in block.run(
+            input_data, user_id=MOCK_USER_ID, credentials=None
+        ):
+            outputs.append((name, data))
+
+        assert len(outputs) == 1
+        assert outputs[0][0] == "result"
+        assert outputs[0][1] == "No auth needed"

autogpt_platform/backend/backend/blocks/mcp/test_mcp.py

@@ -0,0 +1,619 @@
+"""
+Tests for MCP client and MCPToolBlock.
+"""
+
+import json
+from unittest.mock import AsyncMock, patch
+
+import pytest
+
+from backend.blocks.mcp.block import MCPToolBlock
+from backend.blocks.mcp.client import MCPCallResult, MCPClient, MCPClientError
+from backend.util.test import execute_block_test
+
+# ── SSE parsing unit tests ───────────────────────────────────────────
+
+
+class TestSSEParsing:
+    """Tests for SSE (text/event-stream) response parsing."""
+
+    def test_parse_sse_simple(self):
+        sse = (
+            "event: message\n"
+            'data: {"jsonrpc":"2.0","result":{"tools":[]},"id":1}\n'
+            "\n"
+        )
+        body = MCPClient._parse_sse_response(sse)
+        assert body["result"] == {"tools": []}
+        assert body["id"] == 1
+
+    def test_parse_sse_with_notifications(self):
+        """SSE streams can contain notifications (no id) before the response."""
+        sse = (
+            "event: message\n"
+            'data: {"jsonrpc":"2.0","method":"some/notification"}\n'
+            "\n"
+            "event: message\n"
+            'data: {"jsonrpc":"2.0","result":{"ok":true},"id":2}\n'
+            "\n"
+        )
+        body = MCPClient._parse_sse_response(sse)
+        assert body["result"] == {"ok": True}
+        assert body["id"] == 2
+
+    def test_parse_sse_error_response(self):
+        sse = (
+            "event: message\n"
+            'data: {"jsonrpc":"2.0","error":{"code":-32600,"message":"Bad Request"},"id":1}\n'
+        )
+        body = MCPClient._parse_sse_response(sse)
+        assert "error" in body
+        assert body["error"]["code"] == -32600
+
+    def test_parse_sse_no_data_raises(self):
+        with pytest.raises(MCPClientError, match="No JSON-RPC response found"):
+            MCPClient._parse_sse_response("event: message\n\n")
+
+    def test_parse_sse_empty_raises(self):
+        with pytest.raises(MCPClientError, match="No JSON-RPC response found"):
+            MCPClient._parse_sse_response("")
+
+    def test_parse_sse_ignores_non_data_lines(self):
+        sse = (
+            ": comment line\n"
+            "event: message\n"
+            "id: 123\n"
+            'data: {"jsonrpc":"2.0","result":"ok","id":1}\n'
+            "\n"
+        )
+        body = MCPClient._parse_sse_response(sse)
+        assert body["result"] == "ok"
+
+    def test_parse_sse_uses_last_response(self):
+        """If multiple responses exist, use the last one."""
+        sse = (
+            'data: {"jsonrpc":"2.0","result":"first","id":1}\n'
+            "\n"
+            'data: {"jsonrpc":"2.0","result":"second","id":2}\n'
+            "\n"
+        )
+        body = MCPClient._parse_sse_response(sse)
+        assert body["result"] == "second"
+
+
+# ── MCPClient unit tests ─────────────────────────────────────────────
+
+
+class TestMCPClient:
+    """Tests for the MCP HTTP client."""
+
+    def test_build_headers_without_auth(self):
+        client = MCPClient("https://mcp.example.com")
+        headers = client._build_headers()
+        assert "Authorization" not in headers
+        assert headers["Content-Type"] == "application/json"
+
+    def test_build_headers_with_auth(self):
+        client = MCPClient("https://mcp.example.com", auth_token="my-token")
+        headers = client._build_headers()
+        assert headers["Authorization"] == "Bearer my-token"
+
+    def test_build_jsonrpc_request(self):
+        client = MCPClient("https://mcp.example.com")
+        req = client._build_jsonrpc_request("tools/list")
+        assert req["jsonrpc"] == "2.0"
+        assert req["method"] == "tools/list"
+        assert "id" in req
+        assert "params" not in req
+
+    def test_build_jsonrpc_request_with_params(self):
+        client = MCPClient("https://mcp.example.com")
+        req = client._build_jsonrpc_request(
+            "tools/call", {"name": "test", "arguments": {"x": 1}}
+        )
+        assert req["params"] == {"name": "test", "arguments": {"x": 1}}
+
+    def test_request_id_increments(self):
+        client = MCPClient("https://mcp.example.com")
+        req1 = client._build_jsonrpc_request("tools/list")
+        req2 = client._build_jsonrpc_request("tools/list")
+        assert req2["id"] > req1["id"]
+
+    def test_server_url_trailing_slash_stripped(self):
+        client = MCPClient("https://mcp.example.com/mcp/")
+        assert client.server_url == "https://mcp.example.com/mcp"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_send_request_success(self):
+        client = MCPClient("https://mcp.example.com")
+
+        mock_response = AsyncMock()
+        mock_response.json.return_value = {
+            "jsonrpc": "2.0",
+            "result": {"tools": []},
+            "id": 1,
+        }
+
+        with patch.object(client, "_send_request", return_value={"tools": []}):
+            result = await client._send_request("tools/list")
+            assert result == {"tools": []}
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_send_request_error(self):
+        client = MCPClient("https://mcp.example.com")
+
+        async def mock_send(*args, **kwargs):
+            raise MCPClientError("MCP server error [-32600]: Invalid Request")
+
+        with patch.object(client, "_send_request", side_effect=mock_send):
+            with pytest.raises(MCPClientError, match="Invalid Request"):
+                await client._send_request("tools/list")
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_list_tools(self):
+        client = MCPClient("https://mcp.example.com")
+
+        mock_result = {
+            "tools": [
+                {
+                    "name": "get_weather",
+                    "description": "Get current weather for a city",
+                    "inputSchema": {
+                        "type": "object",
+                        "properties": {"city": {"type": "string"}},
+                        "required": ["city"],
+                    },
+                },
+                {
+                    "name": "search",
+                    "description": "Search the web",
+                    "inputSchema": {
+                        "type": "object",
+                        "properties": {"query": {"type": "string"}},
+                        "required": ["query"],
+                    },
+                },
+            ]
+        }
+
+        with patch.object(client, "_send_request", return_value=mock_result):
+            tools = await client.list_tools()
+
+        assert len(tools) == 2
+        assert tools[0].name == "get_weather"
+        assert tools[0].description == "Get current weather for a city"
+        assert tools[0].input_schema["properties"]["city"]["type"] == "string"
+        assert tools[1].name == "search"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_list_tools_empty(self):
+        client = MCPClient("https://mcp.example.com")
+
+        with patch.object(client, "_send_request", return_value={"tools": []}):
+            tools = await client.list_tools()
+
+        assert tools == []
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_list_tools_none_result(self):
+        client = MCPClient("https://mcp.example.com")
+
+        with patch.object(client, "_send_request", return_value=None):
+            tools = await client.list_tools()
+
+        assert tools == []
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_tool_success(self):
+        client = MCPClient("https://mcp.example.com")
+
+        mock_result = {
+            "content": [
+                {"type": "text", "text": json.dumps({"temp": 20, "city": "London"})}
+            ],
+            "isError": False,
+        }
+
+        with patch.object(client, "_send_request", return_value=mock_result):
+            result = await client.call_tool("get_weather", {"city": "London"})
+
+        assert not result.is_error
+        assert len(result.content) == 1
+        assert result.content[0]["type"] == "text"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_tool_error(self):
+        client = MCPClient("https://mcp.example.com")
+
+        mock_result = {
+            "content": [{"type": "text", "text": "City not found"}],
+            "isError": True,
+        }
+
+        with patch.object(client, "_send_request", return_value=mock_result):
+            result = await client.call_tool("get_weather", {"city": "???"})
+
+        assert result.is_error
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_tool_none_result(self):
+        client = MCPClient("https://mcp.example.com")
+
+        with patch.object(client, "_send_request", return_value=None):
+            result = await client.call_tool("get_weather", {"city": "London"})
+
+        assert result.is_error
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_initialize(self):
+        client = MCPClient("https://mcp.example.com")
+
+        mock_result = {
+            "protocolVersion": "2025-03-26",
+            "capabilities": {"tools": {}},
+            "serverInfo": {"name": "test-server", "version": "1.0.0"},
+        }
+
+        with (
+            patch.object(client, "_send_request", return_value=mock_result) as mock_req,
+            patch.object(client, "_send_notification") as mock_notif,
+        ):
+            result = await client.initialize()
+
+        mock_req.assert_called_once()
+        mock_notif.assert_called_once_with("notifications/initialized")
+        assert result["protocolVersion"] == "2025-03-26"
+
+
+# ── MCPToolBlock unit tests ──────────────────────────────────────────
+
+MOCK_USER_ID = "test-user-123"
+
+
+class TestMCPToolBlock:
+    """Tests for the MCPToolBlock."""
+
+    def test_block_instantiation(self):
+        block = MCPToolBlock()
+        assert block.id == "a0a4b1c2-d3e4-4f56-a7b8-c9d0e1f2a3b4"
+        assert block.name == "MCPToolBlock"
+
+    def test_input_schema_has_required_fields(self):
+        block = MCPToolBlock()
+        schema = block.input_schema.jsonschema()
+        props = schema.get("properties", {})
+        assert "server_url" in props
+        assert "selected_tool" in props
+        assert "tool_arguments" in props
+        assert "credentials" in props
+
+    def test_output_schema(self):
+        block = MCPToolBlock()
+        schema = block.output_schema.jsonschema()
+        props = schema.get("properties", {})
+        assert "result" in props
+        assert "error" in props
+
+    def test_get_input_schema_with_tool_schema(self):
+        tool_schema = {
+            "type": "object",
+            "properties": {"query": {"type": "string"}},
+            "required": ["query"],
+        }
+        data = {"tool_input_schema": tool_schema}
+        result = MCPToolBlock.Input.get_input_schema(data)
+        assert result == tool_schema
+
+    def test_get_input_schema_without_tool_schema(self):
+        result = MCPToolBlock.Input.get_input_schema({})
+        assert result == {}
+
+    def test_get_input_defaults(self):
+        data = {"tool_arguments": {"city": "London"}}
+        result = MCPToolBlock.Input.get_input_defaults(data)
+        assert result == {"city": "London"}
+
+    def test_get_missing_input(self):
+        data = {
+            "tool_input_schema": {
+                "type": "object",
+                "properties": {
+                    "city": {"type": "string"},
+                    "units": {"type": "string"},
+                },
+                "required": ["city", "units"],
+            },
+            "tool_arguments": {"city": "London"},
+        }
+        missing = MCPToolBlock.Input.get_missing_input(data)
+        assert missing == {"units"}
+
+    def test_get_missing_input_all_present(self):
+        data = {
+            "tool_input_schema": {
+                "type": "object",
+                "properties": {"city": {"type": "string"}},
+                "required": ["city"],
+            },
+            "tool_arguments": {"city": "London"},
+        }
+        missing = MCPToolBlock.Input.get_missing_input(data)
+        assert missing == set()
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_run_with_mock(self):
+        """Test the block using the built-in test infrastructure."""
+        block = MCPToolBlock()
+        await execute_block_test(block)
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_run_missing_server_url(self):
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url="",
+            selected_tool="test",
+        )
+        outputs = []
+        async for name, data in block.run(input_data, user_id=MOCK_USER_ID):
+            outputs.append((name, data))
+        assert outputs == [("error", "MCP server URL is required")]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_run_missing_tool(self):
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url="https://mcp.example.com/mcp",
+            selected_tool="",
+        )
+        outputs = []
+        async for name, data in block.run(input_data, user_id=MOCK_USER_ID):
+            outputs.append((name, data))
+        assert outputs == [
+            ("error", "No tool selected. Please select a tool from the dropdown.")
+        ]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_run_success(self):
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url="https://mcp.example.com/mcp",
+            selected_tool="get_weather",
+            tool_input_schema={
+                "type": "object",
+                "properties": {"city": {"type": "string"}},
+            },
+            tool_arguments={"city": "London"},
+        )
+
+        async def mock_call(*args, **kwargs):
+            return {"temp": 20, "city": "London"}
+
+        block._call_mcp_tool = mock_call  # type: ignore
+
+        outputs = []
+        async for name, data in block.run(input_data, user_id=MOCK_USER_ID):
+            outputs.append((name, data))
+
+        assert len(outputs) == 1
+        assert outputs[0][0] == "result"
+        assert outputs[0][1] == {"temp": 20, "city": "London"}
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_run_mcp_error(self):
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url="https://mcp.example.com/mcp",
+            selected_tool="bad_tool",
+        )
+
+        async def mock_call(*args, **kwargs):
+            raise MCPClientError("Tool not found")
+
+        block._call_mcp_tool = mock_call  # type: ignore
+
+        outputs = []
+        async for name, data in block.run(input_data, user_id=MOCK_USER_ID):
+            outputs.append((name, data))
+
+        assert outputs[0][0] == "error"
+        assert "Tool not found" in outputs[0][1]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_mcp_tool_parses_json_text(self):
+        block = MCPToolBlock()
+
+        mock_result = MCPCallResult(
+            content=[
+                {"type": "text", "text": '{"temp": 20}'},
+            ],
+            is_error=False,
+        )
+
+        async def mock_init(self):
+            return {}
+
+        async def mock_call(self, name, args):
+            return mock_result
+
+        with (
+            patch.object(MCPClient, "initialize", mock_init),
+            patch.object(MCPClient, "call_tool", mock_call),
+        ):
+            result = await block._call_mcp_tool(
+                "https://mcp.example.com", "test_tool", {}
+            )
+
+        assert result == {"temp": 20}
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_mcp_tool_plain_text(self):
+        block = MCPToolBlock()
+
+        mock_result = MCPCallResult(
+            content=[
+                {"type": "text", "text": "Hello, world!"},
+            ],
+            is_error=False,
+        )
+
+        async def mock_init(self):
+            return {}
+
+        async def mock_call(self, name, args):
+            return mock_result
+
+        with (
+            patch.object(MCPClient, "initialize", mock_init),
+            patch.object(MCPClient, "call_tool", mock_call),
+        ):
+            result = await block._call_mcp_tool(
+                "https://mcp.example.com", "test_tool", {}
+            )
+
+        assert result == "Hello, world!"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_mcp_tool_multiple_content(self):
+        block = MCPToolBlock()
+
+        mock_result = MCPCallResult(
+            content=[
+                {"type": "text", "text": "Part 1"},
+                {"type": "text", "text": '{"part": 2}'},
+            ],
+            is_error=False,
+        )
+
+        async def mock_init(self):
+            return {}
+
+        async def mock_call(self, name, args):
+            return mock_result
+
+        with (
+            patch.object(MCPClient, "initialize", mock_init),
+            patch.object(MCPClient, "call_tool", mock_call),
+        ):
+            result = await block._call_mcp_tool(
+                "https://mcp.example.com", "test_tool", {}
+            )
+
+        assert result == ["Part 1", {"part": 2}]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_mcp_tool_error_result(self):
+        block = MCPToolBlock()
+
+        mock_result = MCPCallResult(
+            content=[{"type": "text", "text": "Something went wrong"}],
+            is_error=True,
+        )
+
+        async def mock_init(self):
+            return {}
+
+        async def mock_call(self, name, args):
+            return mock_result
+
+        with (
+            patch.object(MCPClient, "initialize", mock_init),
+            patch.object(MCPClient, "call_tool", mock_call),
+        ):
+            with pytest.raises(MCPClientError, match="returned an error"):
+                await block._call_mcp_tool("https://mcp.example.com", "test_tool", {})
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_call_mcp_tool_image_content(self):
+        block = MCPToolBlock()
+
+        mock_result = MCPCallResult(
+            content=[
+                {
+                    "type": "image",
+                    "data": "base64data==",
+                    "mimeType": "image/png",
+                }
+            ],
+            is_error=False,
+        )
+
+        async def mock_init(self):
+            return {}
+
+        async def mock_call(self, name, args):
+            return mock_result
+
+        with (
+            patch.object(MCPClient, "initialize", mock_init),
+            patch.object(MCPClient, "call_tool", mock_call),
+        ):
+            result = await block._call_mcp_tool(
+                "https://mcp.example.com", "test_tool", {}
+            )
+
+        assert result == {
+            "type": "image",
+            "data": "base64data==",
+            "mimeType": "image/png",
+        }
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_run_with_credentials(self):
+        """Verify the block uses OAuth2Credentials and passes auth token."""
+        from pydantic import SecretStr
+
+        from backend.data.model import OAuth2Credentials
+
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url="https://mcp.example.com/mcp",
+            selected_tool="test_tool",
+        )
+
+        captured_tokens: list[str | None] = []
+
+        async def mock_call(server_url, tool_name, arguments, auth_token=None):
+            captured_tokens.append(auth_token)
+            return "ok"
+
+        block._call_mcp_tool = mock_call  # type: ignore
+
+        test_creds = OAuth2Credentials(
+            id="cred-123",
+            provider="mcp",
+            access_token=SecretStr("resolved-token"),
+            refresh_token=SecretStr(""),
+            scopes=[],
+            title="Test MCP credential",
+        )
+
+        async for _ in block.run(
+            input_data, user_id=MOCK_USER_ID, credentials=test_creds
+        ):
+            pass
+
+        assert captured_tokens == ["resolved-token"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_run_without_credentials(self):
+        """Verify the block works without credentials (public server)."""
+        block = MCPToolBlock()
+        input_data = MCPToolBlock.Input(
+            server_url="https://mcp.example.com/mcp",
+            selected_tool="test_tool",
+        )
+
+        captured_tokens: list[str | None] = []
+
+        async def mock_call(server_url, tool_name, arguments, auth_token=None):
+            captured_tokens.append(auth_token)
+            return "ok"
+
+        block._call_mcp_tool = mock_call  # type: ignore
+
+        outputs = []
+        async for name, data in block.run(input_data, user_id=MOCK_USER_ID):
+            outputs.append((name, data))
+
+        assert captured_tokens == [None]
+        assert outputs == [("result", "ok")]

autogpt_platform/backend/backend/blocks/mcp/test_oauth.py

@@ -0,0 +1,242 @@
+"""
+Tests for MCP OAuth handler.
+"""
+
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+from pydantic import SecretStr
+
+from backend.blocks.mcp.client import MCPClient
+from backend.blocks.mcp.oauth import MCPOAuthHandler
+from backend.data.model import OAuth2Credentials
+
+
+def _mock_response(json_data: dict, status: int = 200) -> MagicMock:
+    """Create a mock Response with synchronous json() (matching Requests.Response)."""
+    resp = MagicMock()
+    resp.status = status
+    resp.ok = 200 <= status < 300
+    resp.json.return_value = json_data
+    return resp
+
+
+class TestMCPOAuthHandler:
+    """Tests for the MCPOAuthHandler."""
+
+    def _make_handler(self, **overrides) -> MCPOAuthHandler:
+        defaults = {
+            "client_id": "test-client-id",
+            "client_secret": "test-client-secret",
+            "redirect_uri": "https://app.example.com/callback",
+            "authorize_url": "https://auth.example.com/authorize",
+            "token_url": "https://auth.example.com/token",
+        }
+        defaults.update(overrides)
+        return MCPOAuthHandler(**defaults)
+
+    def test_get_login_url_basic(self):
+        handler = self._make_handler()
+        url = handler.get_login_url(
+            scopes=["read", "write"],
+            state="random-state-token",
+            code_challenge="S256-challenge-value",
+        )
+
+        assert "https://auth.example.com/authorize?" in url
+        assert "response_type=code" in url
+        assert "client_id=test-client-id" in url
+        assert "state=random-state-token" in url
+        assert "code_challenge=S256-challenge-value" in url
+        assert "code_challenge_method=S256" in url
+        assert "scope=read+write" in url
+
+    def test_get_login_url_with_resource(self):
+        handler = self._make_handler(resource_url="https://mcp.example.com/mcp")
+        url = handler.get_login_url(
+            scopes=[], state="state", code_challenge="challenge"
+        )
+
+        assert "resource=https" in url
+
+    def test_get_login_url_without_pkce(self):
+        handler = self._make_handler()
+        url = handler.get_login_url(scopes=["read"], state="state", code_challenge=None)
+
+        assert "code_challenge" not in url
+        assert "code_challenge_method" not in url
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_exchange_code_for_tokens(self):
+        handler = self._make_handler()
+
+        resp = _mock_response(
+            {
+                "access_token": "new-access-token",
+                "refresh_token": "new-refresh-token",
+                "expires_in": 3600,
+                "token_type": "Bearer",
+            }
+        )
+
+        with patch("backend.blocks.mcp.oauth.Requests") as MockRequests:
+            instance = MockRequests.return_value
+            instance.post = AsyncMock(return_value=resp)
+
+            creds = await handler.exchange_code_for_tokens(
+                code="auth-code",
+                scopes=["read"],
+                code_verifier="pkce-verifier",
+            )
+
+        assert isinstance(creds, OAuth2Credentials)
+        assert creds.access_token.get_secret_value() == "new-access-token"
+        assert creds.refresh_token is not None
+        assert creds.refresh_token.get_secret_value() == "new-refresh-token"
+        assert creds.scopes == ["read"]
+        assert creds.access_token_expires_at is not None
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_refresh_tokens(self):
+        handler = self._make_handler()
+
+        existing_creds = OAuth2Credentials(
+            id="existing-id",
+            provider="mcp",
+            access_token=SecretStr("old-token"),
+            refresh_token=SecretStr("old-refresh"),
+            scopes=["read"],
+            title="test",
+        )
+
+        resp = _mock_response(
+            {
+                "access_token": "refreshed-token",
+                "refresh_token": "new-refresh",
+                "expires_in": 3600,
+            }
+        )
+
+        with patch("backend.blocks.mcp.oauth.Requests") as MockRequests:
+            instance = MockRequests.return_value
+            instance.post = AsyncMock(return_value=resp)
+
+            refreshed = await handler._refresh_tokens(existing_creds)
+
+        assert refreshed.id == "existing-id"
+        assert refreshed.access_token.get_secret_value() == "refreshed-token"
+        assert refreshed.refresh_token is not None
+        assert refreshed.refresh_token.get_secret_value() == "new-refresh"
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_refresh_tokens_no_refresh_token(self):
+        handler = self._make_handler()
+
+        creds = OAuth2Credentials(
+            provider="mcp",
+            access_token=SecretStr("token"),
+            scopes=["read"],
+            title="test",
+        )
+
+        with pytest.raises(ValueError, match="No refresh token"):
+            await handler._refresh_tokens(creds)
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_revoke_tokens_no_url(self):
+        handler = self._make_handler(revoke_url=None)
+
+        creds = OAuth2Credentials(
+            provider="mcp",
+            access_token=SecretStr("token"),
+            scopes=[],
+            title="test",
+        )
+
+        result = await handler.revoke_tokens(creds)
+        assert result is False
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_revoke_tokens_with_url(self):
+        handler = self._make_handler(revoke_url="https://auth.example.com/revoke")
+
+        creds = OAuth2Credentials(
+            provider="mcp",
+            access_token=SecretStr("token"),
+            scopes=[],
+            title="test",
+        )
+
+        resp = _mock_response({}, status=200)
+
+        with patch("backend.blocks.mcp.oauth.Requests") as MockRequests:
+            instance = MockRequests.return_value
+            instance.post = AsyncMock(return_value=resp)
+
+            result = await handler.revoke_tokens(creds)
+
+        assert result is True
+
+
+class TestMCPClientDiscovery:
+    """Tests for MCPClient OAuth metadata discovery."""
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_discover_auth_found(self):
+        client = MCPClient("https://mcp.example.com/mcp")
+
+        metadata = {
+            "authorization_servers": ["https://auth.example.com"],
+            "resource": "https://mcp.example.com/mcp",
+        }
+
+        resp = _mock_response(metadata, status=200)
+
+        with patch("backend.blocks.mcp.client.Requests") as MockRequests:
+            instance = MockRequests.return_value
+            instance.get = AsyncMock(return_value=resp)
+
+            result = await client.discover_auth()
+
+        assert result is not None
+        assert result["authorization_servers"] == ["https://auth.example.com"]
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_discover_auth_not_found(self):
+        client = MCPClient("https://mcp.example.com/mcp")
+
+        resp = _mock_response({}, status=404)
+
+        with patch("backend.blocks.mcp.client.Requests") as MockRequests:
+            instance = MockRequests.return_value
+            instance.get = AsyncMock(return_value=resp)
+
+            result = await client.discover_auth()
+
+        assert result is None
+
+    @pytest.mark.asyncio(loop_scope="session")
+    async def test_discover_auth_server_metadata(self):
+        client = MCPClient("https://mcp.example.com/mcp")
+
+        server_metadata = {
+            "issuer": "https://auth.example.com",
+            "authorization_endpoint": "https://auth.example.com/authorize",
+            "token_endpoint": "https://auth.example.com/token",
+            "registration_endpoint": "https://auth.example.com/register",
+            "code_challenge_methods_supported": ["S256"],
+        }
+
+        resp = _mock_response(server_metadata, status=200)
+
+        with patch("backend.blocks.mcp.client.Requests") as MockRequests:
+            instance = MockRequests.return_value
+            instance.get = AsyncMock(return_value=resp)
+
+            result = await client.discover_auth_server_metadata(
+                "https://auth.example.com"
+            )
+
+        assert result is not None
+        assert result["authorization_endpoint"] == "https://auth.example.com/authorize"
+        assert result["token_endpoint"] == "https://auth.example.com/token"

autogpt_platform/backend/backend/blocks/mcp/test_server.py

@@ -0,0 +1,162 @@
+"""
+Minimal MCP server for integration testing.
+
+Implements the MCP Streamable HTTP transport (JSON-RPC 2.0 over HTTP POST)
+with a few sample tools. Runs on localhost with a random available port.
+"""
+
+import json
+import logging
+
+from aiohttp import web
+
+logger = logging.getLogger(__name__)
+
+# Sample tools this test server exposes
+TEST_TOOLS = [
+    {
+        "name": "get_weather",
+        "description": "Get current weather for a city",
+        "inputSchema": {
+            "type": "object",
+            "properties": {
+                "city": {
+                    "type": "string",
+                    "description": "City name",
+                },
+            },
+            "required": ["city"],
+        },
+    },
+    {
+        "name": "add_numbers",
+        "description": "Add two numbers together",
+        "inputSchema": {
+            "type": "object",
+            "properties": {
+                "a": {"type": "number", "description": "First number"},
+                "b": {"type": "number", "description": "Second number"},
+            },
+            "required": ["a", "b"],
+        },
+    },
+    {
+        "name": "echo",
+        "description": "Echo back the input message",
+        "inputSchema": {
+            "type": "object",
+            "properties": {
+                "message": {"type": "string", "description": "Message to echo"},
+            },
+            "required": ["message"],
+        },
+    },
+]
+
+
+def _handle_initialize(params: dict) -> dict:
+    return {
+        "protocolVersion": "2025-03-26",
+        "capabilities": {"tools": {"listChanged": False}},
+        "serverInfo": {"name": "test-mcp-server", "version": "1.0.0"},
+    }
+
+
+def _handle_tools_list(params: dict) -> dict:
+    return {"tools": TEST_TOOLS}
+
+
+def _handle_tools_call(params: dict) -> dict:
+    tool_name = params.get("name", "")
+    arguments = params.get("arguments", {})
+
+    if tool_name == "get_weather":
+        city = arguments.get("city", "Unknown")
+        return {
+            "content": [
+                {
+                    "type": "text",
+                    "text": json.dumps(
+                        {"city": city, "temperature": 22, "condition": "sunny"}
+                    ),
+                }
+            ],
+        }
+
+    elif tool_name == "add_numbers":
+        a = arguments.get("a", 0)
+        b = arguments.get("b", 0)
+        return {
+            "content": [{"type": "text", "text": json.dumps({"result": a + b})}],
+        }
+
+    elif tool_name == "echo":
+        message = arguments.get("message", "")
+        return {
+            "content": [{"type": "text", "text": message}],
+        }
+
+    else:
+        return {
+            "content": [{"type": "text", "text": f"Unknown tool: {tool_name}"}],
+            "isError": True,
+        }
+
+
+HANDLERS = {
+    "initialize": _handle_initialize,
+    "tools/list": _handle_tools_list,
+    "tools/call": _handle_tools_call,
+}
+
+
+async def handle_mcp_request(request: web.Request) -> web.Response:
+    """Handle incoming MCP JSON-RPC 2.0 requests."""
+    # Check auth if configured
+    expected_token = request.app.get("auth_token")
+    if expected_token:
+        auth_header = request.headers.get("Authorization", "")
+        if auth_header != f"Bearer {expected_token}":
+            return web.json_response(
+                {
+                    "jsonrpc": "2.0",
+                    "error": {"code": -32001, "message": "Unauthorized"},
+                    "id": None,
+                },
+                status=401,
+            )
+
+    body = await request.json()
+
+    # Handle notifications (no id field) — just acknowledge
+    if "id" not in body:
+        return web.Response(status=202)
+
+    method = body.get("method", "")
+    params = body.get("params", {})
+    request_id = body.get("id")
+
+    handler = HANDLERS.get(method)
+    if not handler:
+        return web.json_response(
+            {
+                "jsonrpc": "2.0",
+                "error": {
+                    "code": -32601,
+                    "message": f"Method not found: {method}",
+                },
+                "id": request_id,
+            }
+        )
+
+    result = handler(params)
+    return web.json_response({"jsonrpc": "2.0", "result": result, "id": request_id})
+
+
+def create_test_mcp_app(auth_token: str | None = None) -> web.Application:
+    """Create an aiohttp app that acts as an MCP server."""
+    app = web.Application()
+    app.router.add_post("/mcp", handle_mcp_request)
+    if auth_token:
+        app["auth_token"] = auth_token
+    return app

autogpt_platform/backend/backend/blocks/smart_decision_maker.py

@@ -226,10 +226,9 @@ class SmartDecisionMakerBlock(Block):
         )
         model: llm.LlmModel = SchemaField(
             title="LLM Model",
-            default_factory=llm.LlmModel.default,
+            default=llm.DEFAULT_LLM_MODEL,
             description="The language model to use for answering the prompt.",
             advanced=False,
-            json_schema_extra=llm.llm_model_schema_extra(),
         )
         credentials: llm.AICredentials = llm.AICredentialsField()
         multiple_tool_calls: bool = SchemaField(

autogpt_platform/backend/backend/blocks/stagehand/blocks.py

@@ -10,13 +10,13 @@ import stagehand.main
 from stagehand import Stagehand
 
 from backend.blocks.llm import (
+    MODEL_METADATA,
     AICredentials,
     AICredentialsField,
     LlmModel,
     ModelMetadata,
 )
 from backend.blocks.stagehand._config import stagehand as stagehand_provider
-from backend.data import llm_registry
 from backend.sdk import (
     APIKeyCredentials,
     Block,
@@ -91,7 +91,7 @@ class StagehandRecommendedLlmModel(str, Enum):
         Returns the provider name for the model in the required format for Stagehand:
         provider/model_name
         """
-        model_metadata = self.metadata
+        model_metadata = MODEL_METADATA[LlmModel(self.value)]
         model_name = self.value
 
         if len(model_name.split("/")) == 1 and not self.value.startswith(
@@ -102,28 +102,24 @@ class StagehandRecommendedLlmModel(str, Enum):
             ), "Logic failed and open_router provider attempted to be prepended to model name! in stagehand/_config.py"
             model_name = f"{model_metadata.provider}/{model_name}"
 
-        logger.debug(f"Model name: {model_name}")
+        logger.error(f"Model name: {model_name}")
         return model_name
 
     @property
     def provider(self) -> str:
-        return self.metadata.provider
+        return MODEL_METADATA[LlmModel(self.value)].provider
 
     @property
     def metadata(self) -> ModelMetadata:
-        metadata = llm_registry.get_llm_model_metadata(self.value)
-        if metadata:
-            return metadata
-        # Fallback to LlmModel enum if registry lookup fails
-        return LlmModel(self.value).metadata
+        return MODEL_METADATA[LlmModel(self.value)]
 
     @property
     def context_window(self) -> int:
-        return self.metadata.context_window
+        return MODEL_METADATA[LlmModel(self.value)].context_window
 
     @property
     def max_output_tokens(self) -> int | None:
-        return self.metadata.max_output_tokens
+        return MODEL_METADATA[LlmModel(self.value)].max_output_tokens
 
 
 class StagehandObserveBlock(Block):

autogpt_platform/backend/backend/data/block.py

@@ -19,30 +19,6 @@ CompletedBlockOutput = dict[str, list[Any]]  # Completed stream, collected as a
 
 
 async def initialize_blocks() -> None:
-    # Refresh LLM registry before initializing blocks so blocks can use registry data
-    # This ensures the registry cache is populated even in executor context
-    try:
-        from backend.data import llm_registry
-        from backend.data.block_cost_config import refresh_llm_costs
-
-        # Only refresh if we have DB access (check if Prisma is connected)
-        from backend.data.db import is_connected
-
-        if is_connected():
-            await llm_registry.refresh_llm_registry()
-            await refresh_llm_costs()
-            logger.info("LLM registry refreshed during block initialization")
-        else:
-            logger.warning(
-                "Prisma not connected, skipping LLM registry refresh during block initialization"
-            )
-    except Exception as exc:
-        logger.warning(
-            "Failed to refresh LLM registry during block initialization: %s", exc
-        )
-
-    # First, sync all provider costs to blocks
-    # Imported here to avoid circular import
     from backend.blocks import get_blocks
     from backend.sdk.cost_integration import sync_all_provider_costs
     from backend.util.retry import func_retry

autogpt_platform/backend/backend/data/block_cost_config.py

@@ -1,8 +1,5 @@
-import logging
 from typing import Type
 
-import prisma.models
-
 from backend.blocks._base import Block, BlockCost, BlockCostType
 from backend.blocks.ai_image_customizer import AIImageCustomizerBlock, GeminiImageModel
 from backend.blocks.ai_image_generator_block import AIImageGeneratorBlock, ImageGenModel
@@ -27,11 +24,13 @@ from backend.blocks.ideogram import IdeogramModelBlock
 from backend.blocks.jina.embeddings import JinaEmbeddingBlock
 from backend.blocks.jina.search import ExtractWebsiteContentBlock, SearchTheWebBlock
 from backend.blocks.llm import (
+    MODEL_METADATA,
     AIConversationBlock,
     AIListGeneratorBlock,
     AIStructuredResponseGeneratorBlock,
     AITextGeneratorBlock,
     AITextSummarizerBlock,
+    LlmModel,
 )
 from backend.blocks.replicate.flux_advanced import ReplicateFluxAdvancedModelBlock
 from backend.blocks.replicate.replicate_block import ReplicateModelBlock
@@ -39,7 +38,6 @@ from backend.blocks.smart_decision_maker import SmartDecisionMakerBlock
 from backend.blocks.talking_head import CreateTalkingAvatarVideoBlock
 from backend.blocks.text_to_speech_block import UnrealTextToSpeechBlock
 from backend.blocks.video.narration import VideoNarrationBlock
-from backend.data import llm_registry
 from backend.integrations.credentials_store import (
     aiml_api_credentials,
     anthropic_credentials,
@@ -59,116 +57,210 @@ from backend.integrations.credentials_store import (
     v0_credentials,
 )
 
-logger = logging.getLogger(__name__)
+# =============== Configure the cost for each LLM Model call =============== #
 
-PROVIDER_CREDENTIALS = {
-    "openai": openai_credentials,
-    "anthropic": anthropic_credentials,
-    "groq": groq_credentials,
-    "open_router": open_router_credentials,
-    "llama_api": llama_api_credentials,
-    "aiml_api": aiml_api_credentials,
-    "v0": v0_credentials,
+MODEL_COST: dict[LlmModel, int] = {
+    LlmModel.O3: 4,
+    LlmModel.O3_MINI: 2,
+    LlmModel.O1: 16,
+    LlmModel.O1_MINI: 4,
+    # GPT-5 models
+    LlmModel.GPT5_2: 6,
+    LlmModel.GPT5_1: 5,
+    LlmModel.GPT5: 2,
+    LlmModel.GPT5_MINI: 1,
+    LlmModel.GPT5_NANO: 1,
+    LlmModel.GPT5_CHAT: 5,
+    LlmModel.GPT41: 2,
+    LlmModel.GPT41_MINI: 1,
+    LlmModel.GPT4O_MINI: 1,
+    LlmModel.GPT4O: 3,
+    LlmModel.GPT4_TURBO: 10,
+    LlmModel.GPT3_5_TURBO: 1,
+    LlmModel.CLAUDE_4_1_OPUS: 21,
+    LlmModel.CLAUDE_4_OPUS: 21,
+    LlmModel.CLAUDE_4_SONNET: 5,
+    LlmModel.CLAUDE_4_6_OPUS: 14,
+    LlmModel.CLAUDE_4_5_HAIKU: 4,
+    LlmModel.CLAUDE_4_5_OPUS: 14,
+    LlmModel.CLAUDE_4_5_SONNET: 9,
+    LlmModel.CLAUDE_3_HAIKU: 1,
+    LlmModel.AIML_API_QWEN2_5_72B: 1,
+    LlmModel.AIML_API_LLAMA3_1_70B: 1,
+    LlmModel.AIML_API_LLAMA3_3_70B: 1,
+    LlmModel.AIML_API_META_LLAMA_3_1_70B: 1,
+    LlmModel.AIML_API_LLAMA_3_2_3B: 1,
+    LlmModel.LLAMA3_3_70B: 1,
+    LlmModel.LLAMA3_1_8B: 1,
+    LlmModel.OLLAMA_LLAMA3_3: 1,
+    LlmModel.OLLAMA_LLAMA3_2: 1,
+    LlmModel.OLLAMA_LLAMA3_8B: 1,
+    LlmModel.OLLAMA_LLAMA3_405B: 1,
+    LlmModel.OLLAMA_DOLPHIN: 1,
+    LlmModel.OPENAI_GPT_OSS_120B: 1,
+    LlmModel.OPENAI_GPT_OSS_20B: 1,
+    LlmModel.GEMINI_2_5_PRO: 4,
+    LlmModel.GEMINI_3_PRO_PREVIEW: 5,
+    LlmModel.GEMINI_2_5_FLASH: 1,
+    LlmModel.GEMINI_2_0_FLASH: 1,
+    LlmModel.GEMINI_2_5_FLASH_LITE_PREVIEW: 1,
+    LlmModel.GEMINI_2_0_FLASH_LITE: 1,
+    LlmModel.MISTRAL_NEMO: 1,
+    LlmModel.COHERE_COMMAND_R_08_2024: 1,
+    LlmModel.COHERE_COMMAND_R_PLUS_08_2024: 3,
+    LlmModel.DEEPSEEK_CHAT: 2,
+    LlmModel.DEEPSEEK_R1_0528: 1,
+    LlmModel.PERPLEXITY_SONAR: 1,
+    LlmModel.PERPLEXITY_SONAR_PRO: 5,
+    LlmModel.PERPLEXITY_SONAR_DEEP_RESEARCH: 10,
+    LlmModel.NOUSRESEARCH_HERMES_3_LLAMA_3_1_405B: 1,
+    LlmModel.NOUSRESEARCH_HERMES_3_LLAMA_3_1_70B: 1,
+    LlmModel.AMAZON_NOVA_LITE_V1: 1,
+    LlmModel.AMAZON_NOVA_MICRO_V1: 1,
+    LlmModel.AMAZON_NOVA_PRO_V1: 1,
+    LlmModel.MICROSOFT_WIZARDLM_2_8X22B: 1,
+    LlmModel.GRYPHE_MYTHOMAX_L2_13B: 1,
+    LlmModel.META_LLAMA_4_SCOUT: 1,
+    LlmModel.META_LLAMA_4_MAVERICK: 1,
+    LlmModel.LLAMA_API_LLAMA_4_SCOUT: 1,
+    LlmModel.LLAMA_API_LLAMA4_MAVERICK: 1,
+    LlmModel.LLAMA_API_LLAMA3_3_8B: 1,
+    LlmModel.LLAMA_API_LLAMA3_3_70B: 1,
+    LlmModel.GROK_4: 9,
+    LlmModel.GROK_4_FAST: 1,
+    LlmModel.GROK_4_1_FAST: 1,
+    LlmModel.GROK_CODE_FAST_1: 1,
+    LlmModel.KIMI_K2: 1,
+    LlmModel.QWEN3_235B_A22B_THINKING: 1,
+    LlmModel.QWEN3_CODER: 9,
+    # v0 by Vercel models
+    LlmModel.V0_1_5_MD: 1,
+    LlmModel.V0_1_5_LG: 2,
+    LlmModel.V0_1_0_MD: 1,
 }
 
-# =============== Configure the cost for each LLM Model call =============== #
-# All LLM costs now come from the database via llm_registry
-
-LLM_COST: list[BlockCost] = []
+for model in LlmModel:
+    if model not in MODEL_COST:
+        raise ValueError(f"Missing MODEL_COST for model: {model}")
 
 
-async def _build_llm_costs_from_registry() -> list[BlockCost]:
-    """
-    Build BlockCost list from all models in the LLM registry.
-
-    This function checks for active model migrations with customCreditCost overrides.
-    When a model has been migrated with a custom price, that price is used instead
-    of the target model's default cost.
-    """
-    # Query active migrations with custom pricing overrides.
-    # Note: LlmModelMigration is system-level data (no userId field) and this function
-    # is only called during app startup and admin operations, so no user ID filter needed.
-    migration_overrides: dict[str, int] = {}
-    try:
-        active_migrations = await prisma.models.LlmModelMigration.prisma().find_many(
-            where={
-                "isReverted": False,
-                "customCreditCost": {"not": None},
-            }
-        )
-        # Key by targetModelSlug since that's the model nodes are now using
-        # after migration. The custom cost applies to the target model.
-        migration_overrides = {
-            migration.targetModelSlug: migration.customCreditCost
-            for migration in active_migrations
-            if migration.customCreditCost is not None
-        }
-        if migration_overrides:
-            logger.info(
-                "Found %d active model migrations with custom pricing overrides",
-                len(migration_overrides),
-            )
-    except Exception as exc:
-        logger.warning(
-            "Failed to query model migration overrides: %s. Proceeding with default costs.",
-            exc,
-            exc_info=True,
-        )
-
-    costs: list[BlockCost] = []
-    for model in llm_registry.iter_dynamic_models():
-        for cost in model.costs:
-            credentials = PROVIDER_CREDENTIALS.get(cost.credential_provider)
-            if not credentials:
-                logger.warning(
-                    "Skipping cost entry for %s due to unknown credentials provider %s",
-                    model.slug,
-                    cost.credential_provider,
-                )
-                continue
-
-            # Check if this model has a custom cost override from migration
-            cost_amount = migration_overrides.get(model.slug, cost.credit_cost)
-
-            if model.slug in migration_overrides:
-                logger.debug(
-                    "Applying custom cost override for model %s: %d credits (default: %d)",
-                    model.slug,
-                    cost_amount,
-                    cost.credit_cost,
-                )
-
-            cost_filter = {
-                "model": model.slug,
+LLM_COST = (
+    # Anthropic Models
+    [
+        BlockCost(
+            cost_type=BlockCostType.RUN,
+            cost_filter={
+                "model": model,
                 "credentials": {
-                    "id": credentials.id,
-                    "provider": credentials.provider,
-                    "type": credentials.type,
+                    "id": anthropic_credentials.id,
+                    "provider": anthropic_credentials.provider,
+                    "type": anthropic_credentials.type,
                 },
-            }
-            costs.append(
-                BlockCost(
-                    cost_type=BlockCostType.RUN,
-                    cost_filter=cost_filter,
-                    cost_amount=cost_amount,
-                )
-            )
-    return costs
-
-
-async def refresh_llm_costs() -> None:
-    """
-    Refresh LLM costs from the registry. All costs now come from the database.
-
-    This function also checks for active model migrations with custom pricing overrides
-    and applies them to ensure accurate billing.
-    """
-    LLM_COST.clear()
-    LLM_COST.extend(await _build_llm_costs_from_registry())
-
-
-# Initial load will happen after registry is refreshed at startup
-# Don't call refresh_llm_costs() here - it will be called after registry refresh
+            },
+            cost_amount=cost,
+        )
+        for model, cost in MODEL_COST.items()
+        if MODEL_METADATA[model].provider == "anthropic"
+    ]
+    # OpenAI Models
+    + [
+        BlockCost(
+            cost_type=BlockCostType.RUN,
+            cost_filter={
+                "model": model,
+                "credentials": {
+                    "id": openai_credentials.id,
+                    "provider": openai_credentials.provider,
+                    "type": openai_credentials.type,
+                },
+            },
+            cost_amount=cost,
+        )
+        for model, cost in MODEL_COST.items()
+        if MODEL_METADATA[model].provider == "openai"
+    ]
+    # Groq Models
+    + [
+        BlockCost(
+            cost_type=BlockCostType.RUN,
+            cost_filter={
+                "model": model,
+                "credentials": {"id": groq_credentials.id},
+            },
+            cost_amount=cost,
+        )
+        for model, cost in MODEL_COST.items()
+        if MODEL_METADATA[model].provider == "groq"
+    ]
+    # Open Router Models
+    + [
+        BlockCost(
+            cost_type=BlockCostType.RUN,
+            cost_filter={
+                "model": model,
+                "credentials": {
+                    "id": open_router_credentials.id,
+                    "provider": open_router_credentials.provider,
+                    "type": open_router_credentials.type,
+                },
+            },
+            cost_amount=cost,
+        )
+        for model, cost in MODEL_COST.items()
+        if MODEL_METADATA[model].provider == "open_router"
+    ]
+    # Llama API Models
+    + [
+        BlockCost(
+            cost_type=BlockCostType.RUN,
+            cost_filter={
+                "model": model,
+                "credentials": {
+                    "id": llama_api_credentials.id,
+                    "provider": llama_api_credentials.provider,
+                    "type": llama_api_credentials.type,
+                },
+            },
+            cost_amount=cost,
+        )
+        for model, cost in MODEL_COST.items()
+        if MODEL_METADATA[model].provider == "llama_api"
+    ]
+    # v0 by Vercel Models
+    + [
+        BlockCost(
+            cost_type=BlockCostType.RUN,
+            cost_filter={
+                "model": model,
+                "credentials": {
+                    "id": v0_credentials.id,
+                    "provider": v0_credentials.provider,
+                    "type": v0_credentials.type,
+                },
+            },
+            cost_amount=cost,
+        )
+        for model, cost in MODEL_COST.items()
+        if MODEL_METADATA[model].provider == "v0"
+    ]
+    # AI/ML Api Models
+    + [
+        BlockCost(
+            cost_type=BlockCostType.RUN,
+            cost_filter={
+                "model": model,
+                "credentials": {
+                    "id": aiml_api_credentials.id,
+                    "provider": aiml_api_credentials.provider,
+                    "type": aiml_api_credentials.type,
+                },
+            },
+            cost_amount=cost,
+        )
+        for model, cost in MODEL_COST.items()
+        if MODEL_METADATA[model].provider == "aiml_api"
+    ]
+)
 
 # =============== This is the exhaustive list of cost for each Block =============== #
 

autogpt_platform/backend/backend/data/graph.py

@@ -33,6 +33,7 @@ from backend.util import type as type_utils
 from backend.util.exceptions import GraphNotAccessibleError, GraphNotInLibraryError
 from backend.util.json import SafeJson
 from backend.util.models import Pagination
+from backend.util.request import parse_url
 
 from .block import BlockInput
 from .db import BaseDbModel
@@ -449,6 +450,9 @@ class GraphModel(Graph, GraphMeta):
                     continue
                 if ProviderName.HTTP in field.provider:
                     continue
+                # MCP credentials are intentionally split by server URL
+                if ProviderName.MCP in field.provider:
+                    continue
 
                 # If this happens, that means a block implementation probably needs
                 # to be updated.
@@ -505,6 +509,18 @@ class GraphModel(Graph, GraphMeta):
                 "required": ["id", "provider", "type"],
             }
 
+            # Add a descriptive display title when URL-based discriminator values
+            # are present (e.g. "mcp.sentry.dev" instead of just "Mcp")
+            if (
+                field_info.discriminator
+                and not field_info.discriminator_mapping
+                and field_info.discriminator_values
+            ):
+                hostnames = sorted(
+                    parse_url(str(v)).netloc for v in field_info.discriminator_values
+                )
+                field_schema["display_name"] = ", ".join(hostnames)
+
             # Add other (optional) field info items
             field_schema.update(
                 field_info.model_dump(
@@ -549,8 +565,17 @@ class GraphModel(Graph, GraphMeta):
 
         for graph in [self] + self.sub_graphs:
             for node in graph.nodes:
-                # Track if this node requires credentials (credentials_optional=False means required)
-                node_required_map[node.id] = not node.credentials_optional
+                # A node's credentials are optional if either:
+                # 1. The node metadata says so (credentials_optional=True), or
+                # 2. All credential fields on the block have defaults (not required by schema)
+                block_required = node.block.input_schema.get_required_fields()
+                creds_required_by_schema = any(
+                    fname in block_required
+                    for fname in node.block.input_schema.get_credentials_fields()
+                )
+                node_required_map[node.id] = (
+                    not node.credentials_optional and creds_required_by_schema
+                )
 
                 for (
                     field_name,
@@ -776,6 +801,19 @@ class GraphModel(Graph, GraphMeta):
                         "'credentials' and `*_credentials` are reserved"
                     )
 
+            # Check custom block-level validation (e.g., MCP dynamic tool arguments).
+            # Blocks can override get_missing_input to report additional missing fields
+            # beyond the standard top-level required fields.
+            if for_run:
+                credential_fields = InputSchema.get_credentials_fields()
+                custom_missing = InputSchema.get_missing_input(node.input_default)
+                for field_name in custom_missing:
+                    if (
+                        field_name not in provided_inputs
+                        and field_name not in credential_fields
+                    ):
+                        node_errors[node.id][field_name] = "This field is required"
+
             # Get input schema properties and check dependencies
             input_fields = InputSchema.model_fields
 
@@ -829,9 +867,67 @@ class GraphModel(Graph, GraphMeta):
 
         return node_errors
 
+    @staticmethod
+    def prune_invalid_links(graph: BaseGraph) -> int:
+        """
+        Remove invalid/orphan links from the graph.
+
+        This removes links that:
+        - Reference non-existent source or sink nodes
+        - Reference invalid block IDs
+
+        Note: Pin name validation is handled separately in _validate_graph_structure.
+
+        Returns the number of links pruned.
+        """
+        node_map = {v.id: v for v in graph.nodes}
+        original_count = len(graph.links)
+        valid_links = []
+
+        for link in graph.links:
+            source_node = node_map.get(link.source_id)
+            sink_node = node_map.get(link.sink_id)
+
+            # Skip if either node doesn't exist
+            if not source_node or not sink_node:
+                logger.warning(
+                    f"Pruning orphan link: source={link.source_id}, sink={link.sink_id} "
+                    f"- node(s) not found"
+                )
+                continue
+
+            # Skip if source block doesn't exist
+            source_block = get_block(source_node.block_id)
+            if not source_block:
+                logger.warning(
+                    f"Pruning link with invalid source block: {source_node.block_id}"
+                )
+                continue
+
+            # Skip if sink block doesn't exist
+            sink_block = get_block(sink_node.block_id)
+            if not sink_block:
+                logger.warning(
+                    f"Pruning link with invalid sink block: {sink_node.block_id}"
+                )
+                continue
+
+            valid_links.append(link)
+
+        graph.links = valid_links
+        pruned_count = original_count - len(valid_links)
+
+        if pruned_count > 0:
+            logger.info(f"Pruned {pruned_count} invalid link(s) from graph {graph.id}")
+
+        return pruned_count
+
     @staticmethod
     def _validate_graph_structure(graph: BaseGraph):
         """Validate graph structure (links, connections, etc.)"""
+        # First, prune invalid links to clean up orphan edges
+        GraphModel.prune_invalid_links(graph)
+
         node_map = {v.id: v for v in graph.nodes}
 
         def is_static_output_block(nid: str) -> bool:
@@ -1625,10 +1721,8 @@ async def migrate_llm_models(migrate_to: LlmModel):
             if field.annotation == LlmModel:
                 llm_model_fields[block.id] = field_name
 
-    # Get all model slugs from the registry (dynamic, not hardcoded enum)
-    from backend.data import llm_registry
-
-    enum_values = list(llm_registry.get_all_model_slugs_for_validation())
+    # Convert enum values to a list of strings for the SQL query
+    enum_values = [v.value for v in LlmModel]
     escaped_enum_values = repr(tuple(enum_values))  # hack but works
 
     # Update each block

autogpt_platform/backend/backend/data/graph_test.py

@@ -462,3 +462,120 @@ def test_node_credentials_optional_with_other_metadata():
     assert node.credentials_optional is True
     assert node.metadata["position"] == {"x": 100, "y": 200}
     assert node.metadata["customized_name"] == "My Custom Node"
+
+
+# ============================================================================
+# Tests for MCP Credential Deduplication
+# ============================================================================
+
+
+def test_mcp_credential_combine_different_servers():
+    """Two MCP credential fields with different server URLs should produce
+    separate entries when combined (not merged into one)."""
+    from backend.data.model import CredentialsFieldInfo, CredentialsType
+    from backend.integrations.providers import ProviderName
+
+    oauth2_types: frozenset[CredentialsType] = frozenset(["oauth2"])
+
+    field_sentry = CredentialsFieldInfo(
+        credentials_provider=frozenset([ProviderName.MCP]),
+        credentials_types=oauth2_types,
+        credentials_scopes=None,
+        discriminator="server_url",
+        discriminator_values={"https://mcp.sentry.dev/mcp"},
+    )
+    field_linear = CredentialsFieldInfo(
+        credentials_provider=frozenset([ProviderName.MCP]),
+        credentials_types=oauth2_types,
+        credentials_scopes=None,
+        discriminator="server_url",
+        discriminator_values={"https://mcp.linear.app/mcp"},
+    )
+
+    combined = CredentialsFieldInfo.combine(
+        (field_sentry, ("node-sentry", "credentials")),
+        (field_linear, ("node-linear", "credentials")),
+    )
+
+    # Should produce 2 separate credential entries
+    assert len(combined) == 2, (
+        f"Expected 2 credential entries for 2 MCP blocks with different servers, "
+        f"got {len(combined)}: {list(combined.keys())}"
+    )
+
+    # Each entry should contain the server hostname in its key
+    keys = list(combined.keys())
+    assert any(
+        "mcp.sentry.dev" in k for k in keys
+    ), f"Expected 'mcp.sentry.dev' in one key, got {keys}"
+    assert any(
+        "mcp.linear.app" in k for k in keys
+    ), f"Expected 'mcp.linear.app' in one key, got {keys}"
+
+
+def test_mcp_credential_combine_same_server():
+    """Two MCP credential fields with the same server URL should be combined
+    into one credential entry."""
+    from backend.data.model import CredentialsFieldInfo, CredentialsType
+    from backend.integrations.providers import ProviderName
+
+    oauth2_types: frozenset[CredentialsType] = frozenset(["oauth2"])
+
+    field_a = CredentialsFieldInfo(
+        credentials_provider=frozenset([ProviderName.MCP]),
+        credentials_types=oauth2_types,
+        credentials_scopes=None,
+        discriminator="server_url",
+        discriminator_values={"https://mcp.sentry.dev/mcp"},
+    )
+    field_b = CredentialsFieldInfo(
+        credentials_provider=frozenset([ProviderName.MCP]),
+        credentials_types=oauth2_types,
+        credentials_scopes=None,
+        discriminator="server_url",
+        discriminator_values={"https://mcp.sentry.dev/mcp"},
+    )
+
+    combined = CredentialsFieldInfo.combine(
+        (field_a, ("node-a", "credentials")),
+        (field_b, ("node-b", "credentials")),
+    )
+
+    # Should produce 1 credential entry (same server URL)
+    assert len(combined) == 1, (
+        f"Expected 1 credential entry for 2 MCP blocks with same server, "
+        f"got {len(combined)}: {list(combined.keys())}"
+    )
+
+
+def test_mcp_credential_combine_no_discriminator_values():
+    """MCP credential fields without discriminator_values should be merged
+    into a single entry (backwards compat for blocks without server_url set)."""
+    from backend.data.model import CredentialsFieldInfo, CredentialsType
+    from backend.integrations.providers import ProviderName
+
+    oauth2_types: frozenset[CredentialsType] = frozenset(["oauth2"])
+
+    field_a = CredentialsFieldInfo(
+        credentials_provider=frozenset([ProviderName.MCP]),
+        credentials_types=oauth2_types,
+        credentials_scopes=None,
+        discriminator="server_url",
+    )
+    field_b = CredentialsFieldInfo(
+        credentials_provider=frozenset([ProviderName.MCP]),
+        credentials_types=oauth2_types,
+        credentials_scopes=None,
+        discriminator="server_url",
+    )
+
+    combined = CredentialsFieldInfo.combine(
+        (field_a, ("node-a", "credentials")),
+        (field_b, ("node-b", "credentials")),
+    )
+
+    # Should produce 1 entry (no URL differentiation)
+    assert len(combined) == 1, (
+        f"Expected 1 credential entry for MCP blocks without discriminator_values, "
+        f"got {len(combined)}: {list(combined.keys())}"
+    )

autogpt_platform/backend/backend/data/llm_registry/__init__.py

@@ -1,72 +0,0 @@
-"""
-LLM Registry module for managing LLM models, providers, and costs dynamically.
-
-This module provides a database-driven registry system for LLM models,
-replacing hardcoded model configurations with a flexible admin-managed system.
-"""
-
-from backend.data.llm_registry.model import ModelMetadata
-
-# Re-export for backwards compatibility
-from backend.data.llm_registry.notifications import (
-    REGISTRY_REFRESH_CHANNEL,
-    publish_registry_refresh_notification,
-    subscribe_to_registry_refresh,
-)
-from backend.data.llm_registry.registry import (
-    RegistryModel,
-    RegistryModelCost,
-    RegistryModelCreator,
-    get_all_model_slugs_for_validation,
-    get_default_model_slug,
-    get_dynamic_model_slugs,
-    get_fallback_model_for_disabled,
-    get_llm_discriminator_mapping,
-    get_llm_model_cost,
-    get_llm_model_metadata,
-    get_llm_model_schema_options,
-    get_model_info,
-    is_model_enabled,
-    iter_dynamic_models,
-    refresh_llm_registry,
-    register_static_costs,
-    register_static_metadata,
-)
-from backend.data.llm_registry.schema_utils import (
-    is_llm_model_field,
-    refresh_llm_discriminator_mapping,
-    refresh_llm_model_options,
-    update_schema_with_llm_registry,
-)
-
-__all__ = [
-    # Types
-    "ModelMetadata",
-    "RegistryModel",
-    "RegistryModelCost",
-    "RegistryModelCreator",
-    # Registry functions
-    "get_all_model_slugs_for_validation",
-    "get_default_model_slug",
-    "get_dynamic_model_slugs",
-    "get_fallback_model_for_disabled",
-    "get_llm_discriminator_mapping",
-    "get_llm_model_cost",
-    "get_llm_model_metadata",
-    "get_llm_model_schema_options",
-    "get_model_info",
-    "is_model_enabled",
-    "iter_dynamic_models",
-    "refresh_llm_registry",
-    "register_static_costs",
-    "register_static_metadata",
-    # Notifications
-    "REGISTRY_REFRESH_CHANNEL",
-    "publish_registry_refresh_notification",
-    "subscribe_to_registry_refresh",
-    # Schema utilities
-    "is_llm_model_field",
-    "refresh_llm_discriminator_mapping",
-    "refresh_llm_model_options",
-    "update_schema_with_llm_registry",
-]

autogpt_platform/backend/backend/data/llm_registry/model.py

@@ -1,25 +0,0 @@
-"""Type definitions for LLM model metadata."""
-
-from typing import Literal, NamedTuple
-
-
-class ModelMetadata(NamedTuple):
-    """Metadata for an LLM model.
-
-    Attributes:
-        provider: The provider identifier (e.g., "openai", "anthropic")
-        context_window: Maximum context window size in tokens
-        max_output_tokens: Maximum output tokens (None if unlimited)
-        display_name: Human-readable name for the model
-        provider_name: Human-readable provider name (e.g., "OpenAI", "Anthropic")
-        creator_name: Name of the organization that created the model
-        price_tier: Relative cost tier (1=cheapest, 2=medium, 3=expensive)
-    """
-
-    provider: str
-    context_window: int
-    max_output_tokens: int | None
-    display_name: str
-    provider_name: str
-    creator_name: str
-    price_tier: Literal[1, 2, 3]

autogpt_platform/backend/backend/data/llm_registry/notifications.py

@@ -1,89 +0,0 @@
-"""
-Redis pub/sub notifications for LLM registry updates.
-
-When models are added/updated/removed via the admin UI, this module
-publishes notifications to Redis that all executor services subscribe to,
-ensuring they refresh their registry cache in real-time.
-"""
-
-import asyncio
-import logging
-from typing import Any
-
-from backend.data.redis_client import connect_async
-
-logger = logging.getLogger(__name__)
-
-# Redis channel name for LLM registry refresh notifications
-REGISTRY_REFRESH_CHANNEL = "llm_registry:refresh"
-
-
-async def publish_registry_refresh_notification() -> None:
-    """
-    Publish a notification to Redis that the LLM registry has been updated.
-    All executor services subscribed to this channel will refresh their registry.
-    """
-    try:
-        redis = await connect_async()
-        await redis.publish(REGISTRY_REFRESH_CHANNEL, "refresh")
-        logger.info("Published LLM registry refresh notification to Redis")
-    except Exception as exc:
-        logger.warning(
-            "Failed to publish LLM registry refresh notification: %s",
-            exc,
-            exc_info=True,
-        )
-
-
-async def subscribe_to_registry_refresh(
-    on_refresh: Any,  # Async callable that takes no args
-) -> None:
-    """
-    Subscribe to Redis notifications for LLM registry updates.
-    This runs in a loop and processes messages as they arrive.
-
-    Args:
-        on_refresh: Async callable to execute when a refresh notification is received
-    """
-    try:
-        redis = await connect_async()
-        pubsub = redis.pubsub()
-        await pubsub.subscribe(REGISTRY_REFRESH_CHANNEL)
-        logger.info(
-            "Subscribed to LLM registry refresh notifications on channel: %s",
-            REGISTRY_REFRESH_CHANNEL,
-        )
-
-        # Process messages in a loop
-        while True:
-            try:
-                message = await pubsub.get_message(
-                    ignore_subscribe_messages=True, timeout=1.0
-                )
-                if (
-                    message
-                    and message["type"] == "message"
-                    and message["channel"] == REGISTRY_REFRESH_CHANNEL
-                ):
-                    logger.info("Received LLM registry refresh notification")
-                    try:
-                        await on_refresh()
-                    except Exception as exc:
-                        logger.error(
-                            "Error refreshing LLM registry from notification: %s",
-                            exc,
-                            exc_info=True,
-                        )
-            except Exception as exc:
-                logger.warning(
-                    "Error processing registry refresh message: %s", exc, exc_info=True
-                )
-                # Continue listening even if one message fails
-                await asyncio.sleep(1)
-    except Exception as exc:
-        logger.error(
-            "Failed to subscribe to LLM registry refresh notifications: %s",
-            exc,
-            exc_info=True,
-        )
-        raise

autogpt_platform/backend/backend/data/llm_registry/registry.py

@@ -1,388 +0,0 @@
-"""Core LLM registry implementation for managing models dynamically."""
-
-from __future__ import annotations
-
-import asyncio
-import logging
-from dataclasses import dataclass, field
-from typing import Any, Iterable
-
-import prisma.models
-
-from backend.data.llm_registry.model import ModelMetadata
-
-logger = logging.getLogger(__name__)
-
-
-def _json_to_dict(value: Any) -> dict[str, Any]:
-    """Convert Prisma Json type to dict, with fallback to empty dict."""
-    if value is None:
-        return {}
-    if isinstance(value, dict):
-        return value
-    # Prisma Json type should always be a dict at runtime
-    return dict(value) if value else {}
-
-
-@dataclass(frozen=True)
-class RegistryModelCost:
-    """Cost configuration for an LLM model."""
-
-    credit_cost: int
-    credential_provider: str
-    credential_id: str | None
-    credential_type: str | None
-    currency: str | None
-    metadata: dict[str, Any]
-
-
-@dataclass(frozen=True)
-class RegistryModelCreator:
-    """Creator information for an LLM model."""
-
-    id: str
-    name: str
-    display_name: str
-    description: str | None
-    website_url: str | None
-    logo_url: str | None
-
-
-@dataclass(frozen=True)
-class RegistryModel:
-    """Represents a model in the LLM registry."""
-
-    slug: str
-    display_name: str
-    description: str | None
-    metadata: ModelMetadata
-    capabilities: dict[str, Any]
-    extra_metadata: dict[str, Any]
-    provider_display_name: str
-    is_enabled: bool
-    is_recommended: bool = False
-    costs: tuple[RegistryModelCost, ...] = field(default_factory=tuple)
-    creator: RegistryModelCreator | None = None
-
-
-_static_metadata: dict[str, ModelMetadata] = {}
-_static_costs: dict[str, int] = {}
-_dynamic_models: dict[str, RegistryModel] = {}
-_schema_options: list[dict[str, str]] = []
-_discriminator_mapping: dict[str, str] = {}
-_lock = asyncio.Lock()
-
-
-def register_static_metadata(metadata: dict[Any, ModelMetadata]) -> None:
-    """Register static metadata for legacy models (deprecated)."""
-    _static_metadata.update({str(key): value for key, value in metadata.items()})
-    _refresh_cached_schema()
-
-
-def register_static_costs(costs: dict[Any, int]) -> None:
-    """Register static costs for legacy models (deprecated)."""
-    _static_costs.update({str(key): value for key, value in costs.items()})
-
-
-def _build_schema_options() -> list[dict[str, str]]:
-    """Build schema options for model selection dropdown. Only includes enabled models."""
-    options: list[dict[str, str]] = []
-    # Only include enabled models in the dropdown options
-    for model in sorted(_dynamic_models.values(), key=lambda m: m.display_name.lower()):
-        if model.is_enabled:
-            options.append(
-                {
-                    "label": model.display_name,
-                    "value": model.slug,
-                    "group": model.metadata.provider,
-                    "description": model.description or "",
-                }
-            )
-
-    for slug, metadata in _static_metadata.items():
-        if slug in _dynamic_models:
-            continue
-        options.append(
-            {
-                "label": slug,
-                "value": slug,
-                "group": metadata.provider,
-                "description": "",
-            }
-        )
-    return options
-
-
-async def refresh_llm_registry() -> None:
-    """Refresh the LLM registry from the database. Loads all models (enabled and disabled)."""
-    async with _lock:
-        try:
-            records = await prisma.models.LlmModel.prisma().find_many(
-                include={
-                    "Provider": True,
-                    "Costs": True,
-                    "Creator": True,
-                }
-            )
-            logger.debug("Found %d LLM model records in database", len(records))
-        except Exception as exc:
-            logger.error(
-                "Failed to refresh LLM registry from DB: %s", exc, exc_info=True
-            )
-            return
-
-        dynamic: dict[str, RegistryModel] = {}
-        for record in records:
-            provider_name = (
-                record.Provider.name if record.Provider else record.providerId
-            )
-            provider_display_name = (
-                record.Provider.displayName if record.Provider else record.providerId
-            )
-            # Creator name: prefer Creator.name, fallback to provider display name
-            creator_name = (
-                record.Creator.name if record.Creator else provider_display_name
-            )
-            # Price tier: default to 1 (cheapest) if not set
-            price_tier = getattr(record, "priceTier", 1) or 1
-            # Clamp to valid range 1-3
-            price_tier = max(1, min(3, price_tier))
-
-            metadata = ModelMetadata(
-                provider=provider_name,
-                context_window=record.contextWindow,
-                max_output_tokens=record.maxOutputTokens,
-                display_name=record.displayName,
-                provider_name=provider_display_name,
-                creator_name=creator_name,
-                price_tier=price_tier,  # type: ignore[arg-type]
-            )
-            costs = tuple(
-                RegistryModelCost(
-                    credit_cost=cost.creditCost,
-                    credential_provider=cost.credentialProvider,
-                    credential_id=cost.credentialId,
-                    credential_type=cost.credentialType,
-                    currency=cost.currency,
-                    metadata=_json_to_dict(cost.metadata),
-                )
-                for cost in (record.Costs or [])
-            )
-
-            # Map creator if present
-            creator = None
-            if record.Creator:
-                creator = RegistryModelCreator(
-                    id=record.Creator.id,
-                    name=record.Creator.name,
-                    display_name=record.Creator.displayName,
-                    description=record.Creator.description,
-                    website_url=record.Creator.websiteUrl,
-                    logo_url=record.Creator.logoUrl,
-                )
-
-            dynamic[record.slug] = RegistryModel(
-                slug=record.slug,
-                display_name=record.displayName,
-                description=record.description,
-                metadata=metadata,
-                capabilities=_json_to_dict(record.capabilities),
-                extra_metadata=_json_to_dict(record.metadata),
-                provider_display_name=(
-                    record.Provider.displayName
-                    if record.Provider
-                    else record.providerId
-                ),
-                is_enabled=record.isEnabled,
-                is_recommended=record.isRecommended,
-                costs=costs,
-                creator=creator,
-            )
-
-        # Atomic swap - build new structures then replace references
-        # This ensures readers never see partially updated state
-        global _dynamic_models
-        _dynamic_models = dynamic
-        _refresh_cached_schema()
-        logger.info(
-            "LLM registry refreshed with %s dynamic models (enabled: %s, disabled: %s)",
-            len(dynamic),
-            sum(1 for m in dynamic.values() if m.is_enabled),
-            sum(1 for m in dynamic.values() if not m.is_enabled),
-        )
-
-
-def _refresh_cached_schema() -> None:
-    """Refresh cached schema options and discriminator mapping."""
-    global _schema_options, _discriminator_mapping
-
-    # Build new structures
-    new_options = _build_schema_options()
-    new_mapping = {
-        slug: entry.metadata.provider for slug, entry in _dynamic_models.items()
-    }
-    for slug, metadata in _static_metadata.items():
-        new_mapping.setdefault(slug, metadata.provider)
-
-    # Atomic swap - replace references to ensure readers see consistent state
-    _schema_options = new_options
-    _discriminator_mapping = new_mapping
-
-
-def get_llm_model_metadata(slug: str) -> ModelMetadata | None:
-    """Get model metadata by slug. Checks dynamic models first, then static metadata."""
-    if slug in _dynamic_models:
-        return _dynamic_models[slug].metadata
-    return _static_metadata.get(slug)
-
-
-def get_llm_model_cost(slug: str) -> tuple[RegistryModelCost, ...]:
-    """Get model cost configuration by slug."""
-    if slug in _dynamic_models:
-        return _dynamic_models[slug].costs
-    cost_value = _static_costs.get(slug)
-    if cost_value is None:
-        return tuple()
-    return (
-        RegistryModelCost(
-            credit_cost=cost_value,
-            credential_provider="static",
-            credential_id=None,
-            credential_type=None,
-            currency=None,
-            metadata={},
-        ),
-    )
-
-
-def get_llm_model_schema_options() -> list[dict[str, str]]:
-    """
-    Get schema options for LLM model selection dropdown.
-
-    Returns a copy of cached schema options that are refreshed when the registry is
-    updated via refresh_llm_registry() (called on startup and via Redis pub/sub).
-    """
-    # Return a copy to prevent external mutation
-    return list(_schema_options)
-
-
-def get_llm_discriminator_mapping() -> dict[str, str]:
-    """
-    Get discriminator mapping for LLM models.
-
-    Returns a copy of cached discriminator mapping that is refreshed when the registry
-    is updated via refresh_llm_registry() (called on startup and via Redis pub/sub).
-    """
-    # Return a copy to prevent external mutation
-    return dict(_discriminator_mapping)
-
-
-def get_dynamic_model_slugs() -> set[str]:
-    """Get all dynamic model slugs from the registry."""
-    return set(_dynamic_models.keys())
-
-
-def get_all_model_slugs_for_validation() -> set[str]:
-    """
-    Get ALL model slugs (both enabled and disabled) for validation purposes.
-
-    This is used for JSON schema enum validation - we need to accept any known
-    model value (even disabled ones) so that existing graphs don't fail validation.
-    The actual fallback/enforcement happens at runtime in llm_call().
-    """
-    all_slugs = set(_dynamic_models.keys())
-    all_slugs.update(_static_metadata.keys())
-    return all_slugs
-
-
-def iter_dynamic_models() -> Iterable[RegistryModel]:
-    """Iterate over all dynamic models in the registry."""
-    return tuple(_dynamic_models.values())
-
-
-def get_fallback_model_for_disabled(disabled_model_slug: str) -> RegistryModel | None:
-    """
-    Find a fallback model when the requested model is disabled.
-
-    Looks for an enabled model from the same provider. Prefers models with
-    similar names or capabilities if possible.
-
-    Args:
-        disabled_model_slug: The slug of the disabled model
-
-    Returns:
-        An enabled RegistryModel from the same provider, or None if no fallback found
-    """
-    disabled_model = _dynamic_models.get(disabled_model_slug)
-    if not disabled_model:
-        return None
-
-    provider = disabled_model.metadata.provider
-
-    # Find all enabled models from the same provider
-    candidates = [
-        model
-        for model in _dynamic_models.values()
-        if model.is_enabled and model.metadata.provider == provider
-    ]
-
-    if not candidates:
-        return None
-
-    # Sort by: prefer models with similar context window, then by name
-    candidates.sort(
-        key=lambda m: (
-            abs(m.metadata.context_window - disabled_model.metadata.context_window),
-            m.display_name.lower(),
-        )
-    )
-
-    return candidates[0]
-
-
-def is_model_enabled(model_slug: str) -> bool:
-    """Check if a model is enabled in the registry."""
-    model = _dynamic_models.get(model_slug)
-    if not model:
-        # Model not in registry - assume it's a static/legacy model and allow it
-        return True
-    return model.is_enabled
-
-
-def get_model_info(model_slug: str) -> RegistryModel | None:
-    """Get model info from the registry."""
-    return _dynamic_models.get(model_slug)
-
-
-def get_default_model_slug() -> str | None:
-    """
-    Get the default model slug to use for block defaults.
-
-    Returns the recommended model if set (configured via admin UI),
-    otherwise returns the first enabled model alphabetically.
-    Returns None if no models are available or enabled.
-    """
-    # Return the recommended model if one is set and enabled
-    for model in _dynamic_models.values():
-        if model.is_recommended and model.is_enabled:
-            return model.slug
-
-    # No recommended model set - find first enabled model alphabetically
-    for model in sorted(_dynamic_models.values(), key=lambda m: m.display_name.lower()):
-        if model.is_enabled:
-            logger.warning(
-                "No recommended model set, using '%s' as default",
-                model.slug,
-            )
-            return model.slug
-
-    # No enabled models available
-    if _dynamic_models:
-        logger.error(
-            "No enabled models found in registry (%d models registered but all disabled)",
-            len(_dynamic_models),
-        )
-    else:
-        logger.error("No models registered in LLM registry")
-
-    return None

autogpt_platform/backend/backend/data/llm_registry/schema_utils.py

@@ -1,130 +0,0 @@
-"""
-Helper utilities for LLM registry integration with block schemas.
-
-This module handles the dynamic injection of discriminator mappings
-and model options from the LLM registry into block schemas.
-"""
-
-import logging
-from typing import Any
-
-from backend.data.llm_registry.registry import (
-    get_all_model_slugs_for_validation,
-    get_default_model_slug,
-    get_llm_discriminator_mapping,
-    get_llm_model_schema_options,
-)
-
-logger = logging.getLogger(__name__)
-
-
-def is_llm_model_field(field_name: str, field_info: Any) -> bool:
-    """
-    Check if a field is an LLM model selection field.
-
-    Returns True if the field has 'options' in json_schema_extra
-    (set by llm_model_schema_extra() in blocks/llm.py).
-    """
-    if not hasattr(field_info, "json_schema_extra"):
-        return False
-
-    extra = field_info.json_schema_extra
-    if isinstance(extra, dict):
-        return "options" in extra
-
-    return False
-
-
-def refresh_llm_model_options(field_schema: dict[str, Any]) -> None:
-    """
-    Refresh LLM model options from the registry.
-
-    Updates 'options' (for frontend dropdown) to show only enabled models,
-    but keeps the 'enum' (for validation) inclusive of ALL known models.
-
-    This is important because:
-    - Options: What users see in the dropdown (enabled models only)
-    - Enum: What values pass validation (all known models, including disabled)
-
-    Existing graphs may have disabled models selected - they should pass validation
-    and the fallback logic in llm_call() will handle using an alternative model.
-    """
-    fresh_options = get_llm_model_schema_options()
-    if not fresh_options:
-        return
-
-    # Update options array (UI dropdown) - only enabled models
-    if "options" in field_schema:
-        field_schema["options"] = fresh_options
-
-    all_known_slugs = get_all_model_slugs_for_validation()
-    if all_known_slugs and "enum" in field_schema:
-        existing_enum = set(field_schema.get("enum", []))
-        combined_enum = existing_enum | all_known_slugs
-        field_schema["enum"] = sorted(combined_enum)
-
-    # Set the default value from the registry (gpt-4o if available, else first enabled)
-    # This ensures new blocks have a sensible default pre-selected
-    default_slug = get_default_model_slug()
-    if default_slug:
-        field_schema["default"] = default_slug
-
-
-def refresh_llm_discriminator_mapping(field_schema: dict[str, Any]) -> None:
-    """
-    Refresh discriminator_mapping for fields that use model-based discrimination.
-
-    The discriminator is already set when AICredentialsField() creates the field.
-    We only need to refresh the mapping when models are added/removed.
-    """
-    if field_schema.get("discriminator") != "model":
-        return
-
-    # Always refresh the mapping to get latest models
-    fresh_mapping = get_llm_discriminator_mapping()
-    if fresh_mapping is not None:
-        field_schema["discriminator_mapping"] = fresh_mapping
-
-
-def update_schema_with_llm_registry(
-    schema: dict[str, Any], model_class: type | None = None
-) -> None:
-    """
-    Update a JSON schema with current LLM registry data.
-
-    Refreshes:
-    1. Model options for LLM model selection fields (dropdown choices)
-    2. Discriminator mappings for credentials fields (model → provider)
-
-    Args:
-        schema: The JSON schema to update (mutated in-place)
-        model_class: The Pydantic model class (optional, for field introspection)
-    """
-    properties = schema.get("properties", {})
-
-    for field_name, field_schema in properties.items():
-        if not isinstance(field_schema, dict):
-            continue
-
-        # Refresh model options for LLM model fields
-        if model_class and hasattr(model_class, "model_fields"):
-            field_info = model_class.model_fields.get(field_name)
-            if field_info and is_llm_model_field(field_name, field_info):
-                try:
-                    refresh_llm_model_options(field_schema)
-                except Exception as exc:
-                    logger.warning(
-                        "Failed to refresh LLM options for field %s: %s",
-                        field_name,
-                        exc,
-                    )
-
-        # Refresh discriminator mapping for fields that use model discrimination
-        try:
-            refresh_llm_discriminator_mapping(field_schema)
-        except Exception as exc:
-            logger.warning(
-                "Failed to refresh discriminator mapping for field %s: %s",
-                field_name,
-                exc,
-            )

autogpt_platform/backend/backend/data/model.py

@@ -29,6 +29,7 @@ from pydantic import (
     GetCoreSchemaHandler,
     SecretStr,
     field_serializer,
+    model_validator,
 )
 from pydantic_core import (
     CoreSchema,
@@ -39,7 +40,6 @@ from pydantic_core import (
 )
 from typing_extensions import TypedDict
 
-from backend.data.llm_registry import update_schema_with_llm_registry
 from backend.integrations.providers import ProviderName
 from backend.util.json import loads as json_loads
 from backend.util.request import parse_url
@@ -503,6 +503,25 @@ class CredentialsMetaInput(BaseModel, Generic[CP, CT]):
     provider: CP
     type: CT
 
+    @model_validator(mode="before")
+    @classmethod
+    def _normalize_legacy_provider(cls, data: Any) -> Any:
+        """Fix ``ProviderName.X`` format from Python 3.13 ``str(Enum)`` bug.
+
+        Python 3.13 changed ``str(StrEnum)`` to return ``"ClassName.MEMBER"``
+        instead of the plain value.  Old stored credential references may have
+        ``provider: "ProviderName.MCP"`` instead of ``"mcp"``.
+        """
+        if isinstance(data, dict):
+            prov = data.get("provider", "")
+            if isinstance(prov, str) and prov.startswith("ProviderName."):
+                member = prov.removeprefix("ProviderName.")
+                try:
+                    data = {**data, "provider": ProviderName[member].value}
+                except KeyError:
+                    pass
+        return data
+
     @classmethod
     def allowed_providers(cls) -> tuple[ProviderName, ...] | None:
         return get_args(cls.model_fields["provider"].annotation)
@@ -551,9 +570,7 @@ class CredentialsMetaInput(BaseModel, Generic[CP, CT]):
             else:
                 schema["credentials_provider"] = allowed_providers
             schema["credentials_types"] = model_class.allowed_cred_types()
-
-        # Ensure LLM discriminators are populated (delegates to shared helper)
-        update_schema_with_llm_registry(schema, model_class)
+        # Do not return anything, just mutate schema in place
 
     model_config = ConfigDict(
         json_schema_extra=_add_json_schema_extra,  # type: ignore
@@ -609,11 +626,18 @@ class CredentialsFieldInfo(BaseModel, Generic[CP, CT]):
         ] = defaultdict(list)
 
         for field, key in fields:
-            if field.provider == frozenset([ProviderName.HTTP]):
-                # HTTP host-scoped credentials can have different hosts that reqires different credential sets.
-                # Group by host extracted from the URL
+            if (
+                field.discriminator
+                and not field.discriminator_mapping
+                and field.discriminator_values
+            ):
+                # URL-based discrimination (e.g. HTTP host-scoped, MCP server URL):
+                # Each unique host gets its own credential entry.
+                provider_prefix = next(iter(field.provider))
+                # Use .value for enum types to get the plain string (e.g. "mcp" not "ProviderName.MCP")
+                prefix_str = getattr(provider_prefix, "value", str(provider_prefix))
                 providers = frozenset(
-                    [cast(CP, "http")]
+                    [cast(CP, prefix_str)]
                     + [
                         cast(CP, parse_url(str(value)).netloc)
                         for value in field.discriminator_values
@@ -708,20 +732,16 @@ def CredentialsField(
     This is enforced by the `BlockSchema` base class.
     """
 
-    # Build field_schema_extra - always include discriminator and mapping if discriminator is set
-    field_schema_extra: dict[str, Any] = {}
-
-    # Always include discriminator if provided
-    if discriminator is not None:
-        field_schema_extra["discriminator"] = discriminator
-        # Always include discriminator_mapping when discriminator is set (even if empty initially)
-        field_schema_extra["discriminator_mapping"] = discriminator_mapping or {}
-
-    # Include other optional fields (only if not None)
-    if required_scopes:
-        field_schema_extra["credentials_scopes"] = list(required_scopes)
-    if discriminator_values:
-        field_schema_extra["discriminator_values"] = discriminator_values
+    field_schema_extra = {
+        k: v
+        for k, v in {
+            "credentials_scopes": list(required_scopes) or None,
+            "discriminator": discriminator,
+            "discriminator_mapping": discriminator_mapping,
+            "discriminator_values": discriminator_values,
+        }.items()
+        if v is not None
+    }
 
     # Merge any json_schema_extra passed in kwargs
     if "json_schema_extra" in kwargs:

autogpt_platform/backend/backend/executor/llm_registry_init.py

@@ -1,67 +0,0 @@
-"""
-Helper functions for LLM registry initialization in executor context.
-
-These functions handle refreshing the LLM registry when the executor starts
-and subscribing to real-time updates via Redis pub/sub.
-"""
-
-import logging
-
-from backend.blocks._base import BlockSchema
-from backend.data import db, llm_registry
-from backend.data.block import initialize_blocks
-from backend.data.block_cost_config import refresh_llm_costs
-from backend.data.llm_registry import subscribe_to_registry_refresh
-
-logger = logging.getLogger(__name__)
-
-
-async def initialize_registry_for_executor() -> None:
-    """
-    Initialize blocks and refresh LLM registry in the executor context.
-
-    This must run in the executor's event loop to have access to the database.
-    """
-    try:
-        # Connect to database if not already connected
-        if not db.is_connected():
-            await db.connect()
-            logger.info("[GraphExecutor] Connected to database for registry refresh")
-
-        # Initialize blocks (internally refreshes LLM registry and costs)
-        await initialize_blocks()
-        logger.info("[GraphExecutor] Blocks initialized")
-    except Exception as exc:
-        logger.warning(
-            "[GraphExecutor] Failed to refresh LLM registry on startup: %s",
-            exc,
-            exc_info=True,
-        )
-
-
-async def refresh_registry_on_notification() -> None:
-    """Refresh LLM registry when notified via Redis pub/sub."""
-    try:
-        # Ensure DB is connected
-        if not db.is_connected():
-            await db.connect()
-
-        # Refresh registry and costs
-        await llm_registry.refresh_llm_registry()
-        await refresh_llm_costs()
-
-        # Clear block schema caches so they regenerate with new model options
-        BlockSchema.clear_all_schema_caches()
-
-        logger.info("[GraphExecutor] LLM registry refreshed from notification")
-    except Exception as exc:
-        logger.error(
-            "[GraphExecutor] Failed to refresh LLM registry from notification: %s",
-            exc,
-            exc_info=True,
-        )
-
-
-async def subscribe_to_registry_updates() -> None:
-    """Subscribe to Redis pub/sub for LLM registry refresh notifications."""
-    await subscribe_to_registry_refresh(refresh_registry_on_notification)

autogpt_platform/backend/backend/executor/manager.py

@@ -20,6 +20,7 @@ from backend.blocks import get_block
 from backend.blocks._base import BlockSchema
 from backend.blocks.agent import AgentExecutorBlock
 from backend.blocks.io import AgentOutputBlock
+from backend.blocks.mcp.block import MCPToolBlock
 from backend.data import redis_client as redis
 from backend.data.block import BlockInput, BlockOutput, BlockOutputEntry
 from backend.data.credit import UsageTransactionMetadata
@@ -228,6 +229,18 @@ async def execute_node(
             _input_data.nodes_input_masks = nodes_input_masks
         _input_data.user_id = user_id
         input_data = _input_data.model_dump()
+    elif isinstance(node_block, MCPToolBlock):
+        _mcp_data = MCPToolBlock.Input(**node.input_default)
+        # Dynamic tool fields are flattened to top-level by validate_exec
+        # (via get_input_defaults). Collect them back into tool_arguments.
+        tool_schema = _mcp_data.tool_input_schema
+        tool_props = set(tool_schema.get("properties", {}).keys())
+        merged_args = {**_mcp_data.tool_arguments}
+        for key in tool_props:
+            if key in input_data:
+                merged_args[key] = input_data[key]
+        _mcp_data.tool_arguments = merged_args
+        input_data = _mcp_data.model_dump()
     data.inputs = input_data
 
     # Execute the node
@@ -264,8 +277,34 @@ async def execute_node(
 
     # Handle regular credentials fields
     for field_name, input_type in input_model.get_credentials_fields().items():
-        credentials_meta = input_type(**input_data[field_name])
-        credentials, lock = await creds_manager.acquire(user_id, credentials_meta.id)
+        field_value = input_data.get(field_name)
+        if not field_value or (
+            isinstance(field_value, dict) and not field_value.get("id")
+        ):
+            # No credentials configured — nullify so JSON schema validation
+            # doesn't choke on the empty default `{}`.
+            input_data[field_name] = None
+            continue  # Block runs without credentials
+
+        credentials_meta = input_type(**field_value)
+        # Write normalized values back so JSON schema validation also passes
+        # (model_validator may have fixed legacy formats like "ProviderName.MCP")
+        input_data[field_name] = credentials_meta.model_dump(mode="json")
+        try:
+            credentials, lock = await creds_manager.acquire(
+                user_id, credentials_meta.id
+            )
+        except ValueError:
+            # Credential was deleted or doesn't exist.
+            # If the field has a default, run without credentials.
+            if input_model.model_fields[field_name].default is not None:
+                log_metadata.warning(
+                    f"Credentials #{credentials_meta.id} not found, "
+                    "running without (field has default)"
+                )
+                input_data[field_name] = None
+                continue
+            raise
         creds_locks.append(lock)
         extra_exec_kwargs[field_name] = credentials
 
@@ -708,20 +747,6 @@ class ExecutionProcessor:
         )
         self.node_execution_thread.start()
         self.node_evaluation_thread.start()
-
-        # Initialize LLM registry and subscribe to updates
-        from backend.executor.llm_registry_init import (
-            initialize_registry_for_executor,
-            subscribe_to_registry_updates,
-        )
-
-        asyncio.run_coroutine_threadsafe(
-            initialize_registry_for_executor(), self.node_execution_loop
-        )
-        asyncio.run_coroutine_threadsafe(
-            subscribe_to_registry_updates(), self.node_execution_loop
-        )
-
         logger.info(f"[GraphExecutor] {self.tid} started")
 
     @error_logged(swallow=False)

autogpt_platform/backend/backend/executor/utils.py

@@ -260,7 +260,13 @@ async def _validate_node_input_credentials(
         # Track if any credential field is missing for this node
         has_missing_credentials = False
 
+        # A credential field is optional if the node metadata says so, or if
+        # the block schema declares a default for the field.
+        required_fields = block.input_schema.get_required_fields()
+        is_creds_optional = node.credentials_optional
+
         for field_name, credentials_meta_type in credentials_fields.items():
+            field_is_optional = is_creds_optional or field_name not in required_fields
             try:
                 # Check nodes_input_masks first, then input_default
                 field_value = None
@@ -273,7 +279,7 @@ async def _validate_node_input_credentials(
                 elif field_name in node.input_default:
                     # For optional credentials, don't use input_default - treat as missing
                     # This prevents stale credential IDs from failing validation
-                    if node.credentials_optional:
+                    if field_is_optional:
                         field_value = None
                     else:
                         field_value = node.input_default[field_name]
@@ -283,8 +289,8 @@ async def _validate_node_input_credentials(
                     isinstance(field_value, dict) and not field_value.get("id")
                 ):
                     has_missing_credentials = True
-                    # If node has credentials_optional flag, mark for skipping instead of error
-                    if node.credentials_optional:
+                    # If credential field is optional, skip instead of error
+                    if field_is_optional:
                         continue  # Don't add error, will be marked for skip after loop
                     else:
                         credential_errors[node.id][
@@ -334,16 +340,16 @@ async def _validate_node_input_credentials(
                 ] = "Invalid credentials: type/provider mismatch"
                 continue
 
-        # If node has optional credentials and any are missing, mark for skipping
-        # But only if there are no other errors for this node
+        # If node has optional credentials and any are missing, allow running without.
+        # The executor will pass credentials=None to the block's run().
         if (
             has_missing_credentials
-            and node.credentials_optional
+            and is_creds_optional
             and node.id not in credential_errors
         ):
-            nodes_to_skip.add(node.id)
             logger.info(
-                f"Node #{node.id} will be skipped: optional credentials not configured"
+                f"Node #{node.id}: optional credentials not configured, "
+                "running without"
             )
 
     return credential_errors, nodes_to_skip

autogpt_platform/backend/backend/executor/utils_test.py

@@ -495,6 +495,7 @@ async def test_validate_node_input_credentials_returns_nodes_to_skip(
     mock_block.input_schema.get_credentials_fields.return_value = {
         "credentials": mock_credentials_field_type
     }
+    mock_block.input_schema.get_required_fields.return_value = {"credentials"}
     mock_node.block = mock_block
 
     # Create mock graph
@@ -508,8 +509,8 @@ async def test_validate_node_input_credentials_returns_nodes_to_skip(
         nodes_input_masks=None,
     )
 
-    # Node should be in nodes_to_skip, not in errors
-    assert mock_node.id in nodes_to_skip
+    # Node should NOT be in nodes_to_skip (runs without credentials) and not in errors
+    assert mock_node.id not in nodes_to_skip
     assert mock_node.id not in errors
 
 
@@ -535,6 +536,7 @@ async def test_validate_node_input_credentials_required_missing_creds_error(
     mock_block.input_schema.get_credentials_fields.return_value = {
         "credentials": mock_credentials_field_type
     }
+    mock_block.input_schema.get_required_fields.return_value = {"credentials"}
     mock_node.block = mock_block
 
     # Create mock graph

autogpt_platform/backend/backend/integrations/credentials_store.py

@@ -22,6 +22,27 @@ from backend.util.settings import Settings
 
 settings = Settings()
 
+
+def provider_matches(stored: str, expected: str) -> bool:
+    """Compare provider strings, handling Python 3.13 ``str(StrEnum)`` bug.
+
+    On Python 3.13, ``str(ProviderName.MCP)`` returns ``"ProviderName.MCP"``
+    instead of ``"mcp"``.  OAuth states persisted with the buggy format need
+    to match when ``expected`` is the canonical value (e.g. ``"mcp"``).
+    """
+    if stored == expected:
+        return True
+    if stored.startswith("ProviderName."):
+        member = stored.removeprefix("ProviderName.")
+        from backend.integrations.providers import ProviderName
+
+        try:
+            return ProviderName[member].value == expected
+        except KeyError:
+            pass
+    return False
+
+
 # This is an overrride since ollama doesn't actually require an API key, but the creddential system enforces one be attached
 ollama_credentials = APIKeyCredentials(
     id="744fdc56-071a-4761-b5a5-0af0ce10a2b5",
@@ -389,7 +410,7 @@ class IntegrationCredentialsStore:
         self, user_id: str, provider: str
     ) -> list[Credentials]:
         credentials = await self.get_all_creds(user_id)
-        return [c for c in credentials if c.provider == provider]
+        return [c for c in credentials if provider_matches(c.provider, provider)]
 
     async def get_authorized_providers(self, user_id: str) -> list[str]:
         credentials = await self.get_all_creds(user_id)
@@ -485,17 +506,6 @@ class IntegrationCredentialsStore:
         async with self.edit_user_integrations(user_id) as user_integrations:
             user_integrations.oauth_states.append(state)
 
-        async with await self.locked_user_integrations(user_id):
-
-            user_integrations = await self._get_user_integrations(user_id)
-            oauth_states = user_integrations.oauth_states
-            oauth_states.append(state)
-            user_integrations.oauth_states = oauth_states
-
-            await self.db_manager.update_user_integrations(
-                user_id=user_id, data=user_integrations
-            )
-
         return token, code_challenge
 
     def _generate_code_challenge(self) -> tuple[str, str]:
@@ -521,7 +531,7 @@ class IntegrationCredentialsStore:
                     state
                     for state in oauth_states
                     if secrets.compare_digest(state.token, token)
-                    and state.provider == provider
+                    and provider_matches(state.provider, provider)
                     and state.expires_at > now.timestamp()
                 ),
                 None,

autogpt_platform/backend/backend/integrations/creds_manager.py

@@ -9,7 +9,10 @@ from redis.asyncio.lock import Lock as AsyncRedisLock
 
 from backend.data.model import Credentials, OAuth2Credentials
 from backend.data.redis_client import get_redis_async
-from backend.integrations.credentials_store import IntegrationCredentialsStore
+from backend.integrations.credentials_store import (
+    IntegrationCredentialsStore,
+    provider_matches,
+)
 from backend.integrations.oauth import CREDENTIALS_BY_PROVIDER, HANDLERS_BY_NAME
 from backend.integrations.providers import ProviderName
 from backend.util.exceptions import MissingConfigError
@@ -137,7 +140,10 @@ class IntegrationCredentialsManager:
         self, user_id: str, credentials: OAuth2Credentials, lock: bool = True
     ) -> OAuth2Credentials:
         async with self._locked(user_id, credentials.id, "refresh"):
-            oauth_handler = await _get_provider_oauth_handler(credentials.provider)
+            if provider_matches(credentials.provider, ProviderName.MCP.value):
+                oauth_handler = create_mcp_oauth_handler(credentials)
+            else:
+                oauth_handler = await _get_provider_oauth_handler(credentials.provider)
             if oauth_handler.needs_refresh(credentials):
                 logger.debug(
                     f"Refreshing '{credentials.provider}' "
@@ -236,3 +242,31 @@ async def _get_provider_oauth_handler(provider_name_str: str) -> "BaseOAuthHandl
         client_secret=client_secret,
         redirect_uri=f"{frontend_base_url}/auth/integrations/oauth_callback",
     )
+
+
+def create_mcp_oauth_handler(
+    credentials: OAuth2Credentials,
+) -> "BaseOAuthHandler":
+    """Create an MCPOAuthHandler from credential metadata for token refresh.
+
+    MCP OAuth handlers have dynamic endpoints discovered per-server, so they
+    can't be registered as singletons in HANDLERS_BY_NAME. Instead, the handler
+    is reconstructed from metadata stored on the credential during initial auth.
+    """
+    from backend.blocks.mcp.oauth import MCPOAuthHandler
+
+    meta = credentials.metadata or {}
+    token_url = meta.get("mcp_token_url", "")
+    if not token_url:
+        raise ValueError(
+            f"MCP credential {credentials.id} is missing 'mcp_token_url' metadata; "
+            "cannot refresh tokens"
+        )
+    return MCPOAuthHandler(
+        client_id=meta.get("mcp_client_id", ""),
+        client_secret=meta.get("mcp_client_secret", ""),
+        redirect_uri="",  # Not needed for token refresh
+        authorize_url="",  # Not needed for token refresh
+        token_url=token_url,
+        resource_url=meta.get("mcp_resource_url"),
+    )

autogpt_platform/backend/backend/integrations/providers.py

@@ -30,6 +30,7 @@ class ProviderName(str, Enum):
     IDEOGRAM = "ideogram"
     JINA = "jina"
     LLAMA_API = "llama_api"
+    MCP = "mcp"
     MEDIUM = "medium"
     MEM0 = "mem0"
     NOTION = "notion"

autogpt_platform/backend/backend/integrations/webhooks/graph_lifecycle_hooks.py

@@ -51,6 +51,21 @@ async def _on_graph_activate(graph: "BaseGraph | GraphModel", user_id: str):
             if (
                 creds_meta := new_node.input_default.get(creds_field_name)
             ) and not await get_credentials(creds_meta["id"]):
+                # If the credential field is optional (has a default in the
+                # schema, or node metadata marks it optional), clear the stale
+                # reference instead of blocking the save.
+                creds_field_optional = (
+                    new_node.credentials_optional
+                    or creds_field_name not in block_input_schema.get_required_fields()
+                )
+                if creds_field_optional:
+                    new_node.input_default[creds_field_name] = {}
+                    logger.warning(
+                        f"Node #{new_node.id}: cleared stale optional "
+                        f"credentials #{creds_meta['id']} for "
+                        f"'{creds_field_name}'"
+                    )
+                    continue
                 raise ValueError(
                     f"Node #{new_node.id} input '{creds_field_name}' updated with "
                     f"non-existent credentials #{creds_meta['id']}"

autogpt_platform/backend/backend/server/v2/llm/db.py

@@ -1,941 +0,0 @@
-from __future__ import annotations
-
-from typing import Any, Iterable, Sequence, cast
-
-import prisma
-import prisma.models
-
-from backend.data.db import transaction
-from backend.server.v2.llm import model as llm_model
-from backend.util.models import Pagination
-
-
-def _json_dict(value: Any | None) -> dict[str, Any]:
-    if not value:
-        return {}
-    if isinstance(value, dict):
-        return value
-    return {}
-
-
-def _map_cost(record: prisma.models.LlmModelCost) -> llm_model.LlmModelCost:
-    return llm_model.LlmModelCost(
-        id=record.id,
-        unit=record.unit,
-        credit_cost=record.creditCost,
-        credential_provider=record.credentialProvider,
-        credential_id=record.credentialId,
-        credential_type=record.credentialType,
-        currency=record.currency,
-        metadata=_json_dict(record.metadata),
-    )
-
-
-def _map_creator(
-    record: prisma.models.LlmModelCreator,
-) -> llm_model.LlmModelCreator:
-    return llm_model.LlmModelCreator(
-        id=record.id,
-        name=record.name,
-        display_name=record.displayName,
-        description=record.description,
-        website_url=record.websiteUrl,
-        logo_url=record.logoUrl,
-        metadata=_json_dict(record.metadata),
-    )
-
-
-def _map_model(record: prisma.models.LlmModel) -> llm_model.LlmModel:
-    costs = []
-    if record.Costs:
-        costs = [_map_cost(cost) for cost in record.Costs]
-
-    creator = None
-    if hasattr(record, "Creator") and record.Creator:
-        creator = _map_creator(record.Creator)
-
-    return llm_model.LlmModel(
-        id=record.id,
-        slug=record.slug,
-        display_name=record.displayName,
-        description=record.description,
-        provider_id=record.providerId,
-        creator_id=record.creatorId,
-        creator=creator,
-        context_window=record.contextWindow,
-        max_output_tokens=record.maxOutputTokens,
-        is_enabled=record.isEnabled,
-        is_recommended=record.isRecommended,
-        capabilities=_json_dict(record.capabilities),
-        metadata=_json_dict(record.metadata),
-        costs=costs,
-    )
-
-
-def _map_provider(record: prisma.models.LlmProvider) -> llm_model.LlmProvider:
-    models: list[llm_model.LlmModel] = []
-    if record.Models:
-        models = [_map_model(model) for model in record.Models]
-
-    return llm_model.LlmProvider(
-        id=record.id,
-        name=record.name,
-        display_name=record.displayName,
-        description=record.description,
-        default_credential_provider=record.defaultCredentialProvider,
-        default_credential_id=record.defaultCredentialId,
-        default_credential_type=record.defaultCredentialType,
-        supports_tools=record.supportsTools,
-        supports_json_output=record.supportsJsonOutput,
-        supports_reasoning=record.supportsReasoning,
-        supports_parallel_tool=record.supportsParallelTool,
-        metadata=_json_dict(record.metadata),
-        models=models,
-    )
-
-
-async def list_providers(
-    include_models: bool = True, enabled_only: bool = False
-) -> list[llm_model.LlmProvider]:
-    """
-    List all LLM providers.
-
-    Args:
-        include_models: Whether to include models for each provider
-        enabled_only: If True, only include enabled models (for public routes)
-    """
-    include: Any = None
-    if include_models:
-        model_where = {"isEnabled": True} if enabled_only else None
-        include = {
-            "Models": {
-                "include": {"Costs": True, "Creator": True},
-                "where": model_where,
-            }
-        }
-    records = await prisma.models.LlmProvider.prisma().find_many(include=include)
-    return [_map_provider(record) for record in records]
-
-
-async def upsert_provider(
-    request: llm_model.UpsertLlmProviderRequest,
-    provider_id: str | None = None,
-) -> llm_model.LlmProvider:
-    data: Any = {
-        "name": request.name,
-        "displayName": request.display_name,
-        "description": request.description,
-        "defaultCredentialProvider": request.default_credential_provider,
-        "defaultCredentialId": request.default_credential_id,
-        "defaultCredentialType": request.default_credential_type,
-        "supportsTools": request.supports_tools,
-        "supportsJsonOutput": request.supports_json_output,
-        "supportsReasoning": request.supports_reasoning,
-        "supportsParallelTool": request.supports_parallel_tool,
-        "metadata": prisma.Json(request.metadata or {}),
-    }
-    include: Any = {"Models": {"include": {"Costs": True, "Creator": True}}}
-    if provider_id:
-        record = await prisma.models.LlmProvider.prisma().update(
-            where={"id": provider_id},
-            data=data,
-            include=include,
-        )
-    else:
-        record = await prisma.models.LlmProvider.prisma().create(
-            data=data,
-            include=include,
-        )
-    if record is None:
-        raise ValueError("Failed to create/update provider")
-    return _map_provider(record)
-
-
-async def delete_provider(provider_id: str) -> bool:
-    """
-    Delete an LLM provider.
-
-    A provider can only be deleted if it has no associated models.
-    Due to onDelete: Restrict on LlmModel.Provider, the database will
-    block deletion if models exist.
-
-    Args:
-        provider_id: UUID of the provider to delete
-
-    Returns:
-        True if deleted successfully
-
-    Raises:
-        ValueError: If provider not found or has associated models
-    """
-    # Check if provider exists
-    provider = await prisma.models.LlmProvider.prisma().find_unique(
-        where={"id": provider_id},
-        include={"Models": True},
-    )
-    if not provider:
-        raise ValueError(f"Provider with id '{provider_id}' not found")
-
-    # Check if provider has any models
-    model_count = len(provider.Models) if provider.Models else 0
-    if model_count > 0:
-        raise ValueError(
-            f"Cannot delete provider '{provider.displayName}' because it has "
-            f"{model_count} model(s). Delete all models first."
-        )
-
-    # Safe to delete
-    await prisma.models.LlmProvider.prisma().delete(where={"id": provider_id})
-    return True
-
-
-async def list_models(
-    provider_id: str | None = None,
-    enabled_only: bool = False,
-    page: int = 1,
-    page_size: int = 50,
-) -> llm_model.LlmModelsResponse:
-    """
-    List LLM models with pagination.
-
-    Args:
-        provider_id: Optional filter by provider ID
-        enabled_only: If True, only return enabled models (for public routes)
-        page: Page number (1-indexed)
-        page_size: Number of models per page
-    """
-    # Validate pagination inputs to avoid runtime errors
-    if page_size < 1:
-        page_size = 50
-    if page < 1:
-        page = 1
-
-    where: Any = {}
-    if provider_id:
-        where["providerId"] = provider_id
-    if enabled_only:
-        where["isEnabled"] = True
-
-    # Get total count for pagination
-    total_items = await prisma.models.LlmModel.prisma().count(
-        where=where if where else None
-    )
-
-    # Calculate pagination
-    skip = (page - 1) * page_size
-    total_pages = (total_items + page_size - 1) // page_size if total_items > 0 else 0
-
-    records = await prisma.models.LlmModel.prisma().find_many(
-        where=where if where else None,
-        include={"Costs": True, "Creator": True},
-        skip=skip,
-        take=page_size,
-    )
-    models = [_map_model(record) for record in records]
-
-    return llm_model.LlmModelsResponse(
-        models=models,
-        pagination=Pagination(
-            total_items=total_items,
-            total_pages=total_pages,
-            current_page=page,
-            page_size=page_size,
-        ),
-    )
-
-
-def _cost_create_payload(
-    costs: Sequence[llm_model.LlmModelCostInput],
-) -> dict[str, Iterable[dict[str, Any]]]:
-
-    create_items = []
-    for cost in costs:
-        item: dict[str, Any] = {
-            "unit": cost.unit,
-            "creditCost": cost.credit_cost,
-            "credentialProvider": cost.credential_provider,
-        }
-        # Only include optional fields if they have values
-        if cost.credential_id:
-            item["credentialId"] = cost.credential_id
-        if cost.credential_type:
-            item["credentialType"] = cost.credential_type
-        if cost.currency:
-            item["currency"] = cost.currency
-        # Handle metadata - use Prisma Json type
-        if cost.metadata is not None and cost.metadata != {}:
-            item["metadata"] = prisma.Json(cost.metadata)
-        create_items.append(item)
-    return {"create": create_items}
-
-
-async def create_model(
-    request: llm_model.CreateLlmModelRequest,
-) -> llm_model.LlmModel:
-    data: Any = {
-        "slug": request.slug,
-        "displayName": request.display_name,
-        "description": request.description,
-        "Provider": {"connect": {"id": request.provider_id}},
-        "contextWindow": request.context_window,
-        "maxOutputTokens": request.max_output_tokens,
-        "isEnabled": request.is_enabled,
-        "capabilities": prisma.Json(request.capabilities or {}),
-        "metadata": prisma.Json(request.metadata or {}),
-        "Costs": _cost_create_payload(request.costs),
-    }
-    if request.creator_id:
-        data["Creator"] = {"connect": {"id": request.creator_id}}
-
-    record = await prisma.models.LlmModel.prisma().create(
-        data=data,
-        include={"Costs": True, "Creator": True, "Provider": True},
-    )
-    return _map_model(record)
-
-
-async def update_model(
-    model_id: str,
-    request: llm_model.UpdateLlmModelRequest,
-) -> llm_model.LlmModel:
-    # Build scalar field updates (non-relation fields)
-    scalar_data: Any = {}
-    if request.display_name is not None:
-        scalar_data["displayName"] = request.display_name
-    if request.description is not None:
-        scalar_data["description"] = request.description
-    if request.context_window is not None:
-        scalar_data["contextWindow"] = request.context_window
-    if request.max_output_tokens is not None:
-        scalar_data["maxOutputTokens"] = request.max_output_tokens
-    if request.is_enabled is not None:
-        scalar_data["isEnabled"] = request.is_enabled
-    if request.capabilities is not None:
-        scalar_data["capabilities"] = request.capabilities
-    if request.metadata is not None:
-        scalar_data["metadata"] = request.metadata
-    # Foreign keys can be updated directly as scalar fields
-    if request.provider_id is not None:
-        scalar_data["providerId"] = request.provider_id
-    if request.creator_id is not None:
-        # Empty string means remove the creator
-        scalar_data["creatorId"] = request.creator_id if request.creator_id else None
-
-    # If we have costs to update, we need to handle them separately
-    # because nested writes have different constraints
-    if request.costs is not None:
-        # Wrap cost replacement in a transaction for atomicity
-        async with transaction() as tx:
-            # First update scalar fields
-            if scalar_data:
-                await tx.llmmodel.update(
-                    where={"id": model_id},
-                    data=scalar_data,
-                )
-            # Then handle costs: delete existing and create new
-            await tx.llmmodelcost.delete_many(where={"llmModelId": model_id})
-            if request.costs:
-                cost_payload = _cost_create_payload(request.costs)
-                for cost_item in cost_payload["create"]:
-                    cost_item["llmModelId"] = model_id
-                    await tx.llmmodelcost.create(data=cast(Any, cost_item))
-        # Fetch the updated record (outside transaction)
-        record = await prisma.models.LlmModel.prisma().find_unique(
-            where={"id": model_id},
-            include={"Costs": True, "Creator": True},
-        )
-    else:
-        # No costs update - simple update
-        record = await prisma.models.LlmModel.prisma().update(
-            where={"id": model_id},
-            data=scalar_data,
-            include={"Costs": True, "Creator": True},
-        )
-
-    if not record:
-        raise ValueError(f"Model with id '{model_id}' not found")
-    return _map_model(record)
-
-
-async def toggle_model(
-    model_id: str,
-    is_enabled: bool,
-    migrate_to_slug: str | None = None,
-    migration_reason: str | None = None,
-    custom_credit_cost: int | None = None,
-) -> llm_model.ToggleLlmModelResponse:
-    """
-    Toggle a model's enabled status, optionally migrating workflows when disabling.
-
-    Args:
-        model_id: UUID of the model to toggle
-        is_enabled: New enabled status
-        migrate_to_slug: If disabling and this is provided, migrate all workflows
-                         using this model to the specified replacement model
-        migration_reason: Optional reason for the migration (e.g., "Provider outage")
-        custom_credit_cost: Optional custom pricing override for migrated workflows.
-                           When set, the billing system should use this cost instead
-                           of the target model's cost for affected nodes.
-
-    Returns:
-        ToggleLlmModelResponse with the updated model and optional migration stats
-    """
-    import json
-
-    # Get the model being toggled
-    model = await prisma.models.LlmModel.prisma().find_unique(
-        where={"id": model_id}, include={"Costs": True}
-    )
-    if not model:
-        raise ValueError(f"Model with id '{model_id}' not found")
-
-    nodes_migrated = 0
-    migration_id: str | None = None
-
-    # If disabling with migration, perform migration first
-    if not is_enabled and migrate_to_slug:
-        # Validate replacement model exists and is enabled
-        replacement = await prisma.models.LlmModel.prisma().find_unique(
-            where={"slug": migrate_to_slug}
-        )
-        if not replacement:
-            raise ValueError(f"Replacement model '{migrate_to_slug}' not found")
-        if not replacement.isEnabled:
-            raise ValueError(
-                f"Replacement model '{migrate_to_slug}' is disabled. "
-                f"Please enable it before using it as a replacement."
-            )
-
-        # Perform all operations atomically within a single transaction
-        # This ensures no nodes are missed between query and update
-        async with transaction() as tx:
-            # Get the IDs of nodes that will be migrated (inside transaction for consistency)
-            node_ids_result = await tx.query_raw(
-                """
-                SELECT id
-                FROM "AgentNode"
-                WHERE "constantInput"::jsonb->>'model' = $1
-                FOR UPDATE
-                """,
-                model.slug,
-            )
-            migrated_node_ids = (
-                [row["id"] for row in node_ids_result] if node_ids_result else []
-            )
-            nodes_migrated = len(migrated_node_ids)
-
-            if nodes_migrated > 0:
-                # Update by IDs to ensure we only update the exact nodes we queried
-                # Use JSON array and jsonb_array_elements_text for safe parameterization
-                node_ids_json = json.dumps(migrated_node_ids)
-                await tx.execute_raw(
-                    """
-                    UPDATE "AgentNode"
-                    SET "constantInput" = JSONB_SET(
-                        "constantInput"::jsonb,
-                        '{model}',
-                        to_jsonb($1::text)
-                    )
-                    WHERE id::text IN (
-                        SELECT jsonb_array_elements_text($2::jsonb)
-                    )
-                    """,
-                    migrate_to_slug,
-                    node_ids_json,
-                )
-
-            record = await tx.llmmodel.update(
-                where={"id": model_id},
-                data={"isEnabled": is_enabled},
-                include={"Costs": True},
-            )
-
-            # Create migration record for revert capability
-            if nodes_migrated > 0:
-                migration_data: Any = {
-                    "sourceModelSlug": model.slug,
-                    "targetModelSlug": migrate_to_slug,
-                    "reason": migration_reason,
-                    "migratedNodeIds": json.dumps(migrated_node_ids),
-                    "nodeCount": nodes_migrated,
-                    "customCreditCost": custom_credit_cost,
-                }
-                migration_record = await tx.llmmodelmigration.create(
-                    data=migration_data
-                )
-                migration_id = migration_record.id
-    else:
-        # Simple toggle without migration
-        record = await prisma.models.LlmModel.prisma().update(
-            where={"id": model_id},
-            data={"isEnabled": is_enabled},
-            include={"Costs": True},
-        )
-
-    if record is None:
-        raise ValueError(f"Model with id '{model_id}' not found")
-    return llm_model.ToggleLlmModelResponse(
-        model=_map_model(record),
-        nodes_migrated=nodes_migrated,
-        migrated_to_slug=migrate_to_slug if nodes_migrated > 0 else None,
-        migration_id=migration_id,
-    )
-
-
-async def get_model_usage(model_id: str) -> llm_model.LlmModelUsageResponse:
-    """Get usage count for a model."""
-    import prisma as prisma_module
-
-    model = await prisma.models.LlmModel.prisma().find_unique(where={"id": model_id})
-    if not model:
-        raise ValueError(f"Model with id '{model_id}' not found")
-
-    count_result = await prisma_module.get_client().query_raw(
-        """
-        SELECT COUNT(*) as count
-        FROM "AgentNode"
-        WHERE "constantInput"::jsonb->>'model' = $1
-        """,
-        model.slug,
-    )
-    node_count = int(count_result[0]["count"]) if count_result else 0
-
-    return llm_model.LlmModelUsageResponse(model_slug=model.slug, node_count=node_count)
-
-
-async def delete_model(
-    model_id: str, replacement_model_slug: str | None = None
-) -> llm_model.DeleteLlmModelResponse:
-    """
-    Delete a model and optionally migrate all AgentNodes using it to a replacement model.
-
-    This performs an atomic operation within a database transaction:
-    1. Validates the model exists
-    2. Counts affected nodes
-    3. If nodes exist, validates replacement model and migrates them
-    4. Deletes the LlmModel record (CASCADE deletes costs)
-
-    Args:
-        model_id: UUID of the model to delete
-        replacement_model_slug: Slug of the model to migrate to (required only if nodes use this model)
-
-    Returns:
-        DeleteLlmModelResponse with migration stats
-
-    Raises:
-        ValueError: If model not found, nodes exist but no replacement provided,
-                    replacement not found, or replacement is disabled
-    """
-    # 1. Get the model being deleted (early validation - outside transaction)
-    model = await prisma.models.LlmModel.prisma().find_unique(
-        where={"id": model_id}, include={"Costs": True}
-    )
-    if not model:
-        raise ValueError(f"Model with id '{model_id}' not found")
-
-    deleted_slug = model.slug
-    deleted_display_name = model.displayName
-
-    # 2. Perform all mutation logic atomically within a transaction
-    # This prevents TOCTOU issues where nodes could be created between count and delete
-    async with transaction() as tx:
-        # Count affected nodes inside the transaction
-        count_result = await tx.query_raw(
-            """
-            SELECT COUNT(*) as count
-            FROM "AgentNode"
-            WHERE "constantInput"::jsonb->>'model' = $1
-            """,
-            deleted_slug,
-        )
-        nodes_to_migrate = int(count_result[0]["count"]) if count_result else 0
-
-        # Validate replacement model only if there are nodes to migrate
-        if nodes_to_migrate > 0:
-            if not replacement_model_slug:
-                raise ValueError(
-                    f"Cannot delete model '{deleted_slug}': {nodes_to_migrate} workflow node(s) "
-                    f"are using it. Please provide a replacement_model_slug to migrate them."
-                )
-            replacement = await tx.llmmodel.find_unique(
-                where={"slug": replacement_model_slug}
-            )
-            if not replacement:
-                raise ValueError(
-                    f"Replacement model '{replacement_model_slug}' not found"
-                )
-            if not replacement.isEnabled:
-                raise ValueError(
-                    f"Replacement model '{replacement_model_slug}' is disabled. "
-                    f"Please enable it before using it as a replacement."
-                )
-
-            # Migrate all AgentNode.constantInput->model to replacement
-            await tx.execute_raw(
-                """
-                UPDATE "AgentNode"
-                SET "constantInput" = JSONB_SET(
-                    "constantInput"::jsonb,
-                    '{model}',
-                    to_jsonb($1::text)
-                )
-                WHERE "constantInput"::jsonb->>'model' = $2
-                """,
-                replacement_model_slug,
-                deleted_slug,
-            )
-
-        # Delete the model (CASCADE will delete costs automatically)
-        await tx.llmmodel.delete(where={"id": model_id})
-
-    # Build appropriate message based on whether migration happened
-    if nodes_to_migrate > 0:
-        message = (
-            f"Successfully deleted model '{deleted_display_name}' ({deleted_slug}) "
-            f"and migrated {nodes_to_migrate} workflow node(s) to '{replacement_model_slug}'."
-        )
-    else:
-        message = (
-            f"Successfully deleted model '{deleted_display_name}' ({deleted_slug}). "
-            f"No workflows were using this model."
-        )
-
-    return llm_model.DeleteLlmModelResponse(
-        deleted_model_slug=deleted_slug,
-        deleted_model_display_name=deleted_display_name,
-        replacement_model_slug=replacement_model_slug,
-        nodes_migrated=nodes_to_migrate,
-        message=message,
-    )
-
-
-def _map_migration(
-    record: prisma.models.LlmModelMigration,
-) -> llm_model.LlmModelMigration:
-    return llm_model.LlmModelMigration(
-        id=record.id,
-        source_model_slug=record.sourceModelSlug,
-        target_model_slug=record.targetModelSlug,
-        reason=record.reason,
-        node_count=record.nodeCount,
-        custom_credit_cost=record.customCreditCost,
-        is_reverted=record.isReverted,
-        created_at=record.createdAt,
-        reverted_at=record.revertedAt,
-    )
-
-
-async def list_migrations(
-    include_reverted: bool = False,
-) -> list[llm_model.LlmModelMigration]:
-    """
-    List model migrations, optionally including reverted ones.
-
-    Args:
-        include_reverted: If True, include reverted migrations. Default is False.
-
-    Returns:
-        List of LlmModelMigration records
-    """
-    where: Any = None if include_reverted else {"isReverted": False}
-    records = await prisma.models.LlmModelMigration.prisma().find_many(
-        where=where,
-        order={"createdAt": "desc"},
-    )
-    return [_map_migration(record) for record in records]
-
-
-async def get_migration(migration_id: str) -> llm_model.LlmModelMigration | None:
-    """Get a specific migration by ID."""
-    record = await prisma.models.LlmModelMigration.prisma().find_unique(
-        where={"id": migration_id}
-    )
-    return _map_migration(record) if record else None
-
-
-async def revert_migration(
-    migration_id: str,
-    re_enable_source_model: bool = True,
-) -> llm_model.RevertMigrationResponse:
-    """
-    Revert a model migration, restoring affected nodes to their original model.
-
-    This only reverts the specific nodes that were migrated, not all nodes
-    currently using the target model.
-
-    Args:
-        migration_id: UUID of the migration to revert
-        re_enable_source_model: Whether to re-enable the source model if it's disabled
-
-    Returns:
-        RevertMigrationResponse with revert stats
-
-    Raises:
-        ValueError: If migration not found, already reverted, or source model not available
-    """
-    import json
-    from datetime import datetime, timezone
-
-    # Get the migration record
-    migration = await prisma.models.LlmModelMigration.prisma().find_unique(
-        where={"id": migration_id}
-    )
-    if not migration:
-        raise ValueError(f"Migration with id '{migration_id}' not found")
-
-    if migration.isReverted:
-        raise ValueError(
-            f"Migration '{migration_id}' has already been reverted "
-            f"on {migration.revertedAt.isoformat() if migration.revertedAt else 'unknown date'}"
-        )
-
-    # Check if source model exists
-    source_model = await prisma.models.LlmModel.prisma().find_unique(
-        where={"slug": migration.sourceModelSlug}
-    )
-    if not source_model:
-        raise ValueError(
-            f"Source model '{migration.sourceModelSlug}' no longer exists. "
-            f"Cannot revert migration."
-        )
-
-    # Get the migrated node IDs (Prisma auto-parses JSONB to list)
-    migrated_node_ids: list[str] = (
-        migration.migratedNodeIds
-        if isinstance(migration.migratedNodeIds, list)
-        else json.loads(migration.migratedNodeIds)  # type: ignore
-    )
-    if not migrated_node_ids:
-        raise ValueError("No nodes to revert in this migration")
-
-    # Track if we need to re-enable the source model
-    source_model_was_disabled = not source_model.isEnabled
-    should_re_enable = source_model_was_disabled and re_enable_source_model
-    source_model_re_enabled = False
-
-    # Perform revert atomically
-    async with transaction() as tx:
-        # Re-enable the source model if requested and it was disabled
-        if should_re_enable:
-            await tx.llmmodel.update(
-                where={"id": source_model.id},
-                data={"isEnabled": True},
-            )
-            source_model_re_enabled = True
-
-        # Update only the specific nodes that were migrated
-        # We need to check that they still have the target model (haven't been changed since)
-        # Use a single batch update for efficiency
-        # Use JSON array and jsonb_array_elements_text for safe parameterization
-        node_ids_json = json.dumps(migrated_node_ids)
-        result = await tx.execute_raw(
-            """
-            UPDATE "AgentNode"
-            SET "constantInput" = JSONB_SET(
-                "constantInput"::jsonb,
-                '{model}',
-                to_jsonb($1::text)
-            )
-            WHERE id::text IN (
-                SELECT jsonb_array_elements_text($2::jsonb)
-            )
-            AND "constantInput"::jsonb->>'model' = $3
-            """,
-            migration.sourceModelSlug,
-            node_ids_json,
-            migration.targetModelSlug,
-        )
-        nodes_reverted = result if result else 0
-
-        # Mark migration as reverted
-        await tx.llmmodelmigration.update(
-            where={"id": migration_id},
-            data={
-                "isReverted": True,
-                "revertedAt": datetime.now(timezone.utc),
-            },
-        )
-
-    # Calculate nodes that were already changed since migration
-    nodes_already_changed = len(migrated_node_ids) - nodes_reverted
-
-    # Build appropriate message
-    message_parts = [
-        f"Successfully reverted migration: {nodes_reverted} node(s) restored "
-        f"from '{migration.targetModelSlug}' to '{migration.sourceModelSlug}'."
-    ]
-    if nodes_already_changed > 0:
-        message_parts.append(
-            f" {nodes_already_changed} node(s) were already changed and not reverted."
-        )
-    if source_model_re_enabled:
-        message_parts.append(
-            f" Model '{migration.sourceModelSlug}' has been re-enabled."
-        )
-
-    return llm_model.RevertMigrationResponse(
-        migration_id=migration_id,
-        source_model_slug=migration.sourceModelSlug,
-        target_model_slug=migration.targetModelSlug,
-        nodes_reverted=nodes_reverted,
-        nodes_already_changed=nodes_already_changed,
-        source_model_re_enabled=source_model_re_enabled,
-        message="".join(message_parts),
-    )
-
-
-# ============================================================================
-# Creator CRUD operations
-# ============================================================================
-
-
-async def list_creators() -> list[llm_model.LlmModelCreator]:
-    """List all LLM model creators."""
-    records = await prisma.models.LlmModelCreator.prisma().find_many(
-        order={"displayName": "asc"}
-    )
-    return [_map_creator(record) for record in records]
-
-
-async def get_creator(creator_id: str) -> llm_model.LlmModelCreator | None:
-    """Get a specific creator by ID."""
-    record = await prisma.models.LlmModelCreator.prisma().find_unique(
-        where={"id": creator_id}
-    )
-    return _map_creator(record) if record else None
-
-
-async def upsert_creator(
-    request: llm_model.UpsertLlmCreatorRequest,
-    creator_id: str | None = None,
-) -> llm_model.LlmModelCreator:
-    """Create or update a model creator."""
-    data: Any = {
-        "name": request.name,
-        "displayName": request.display_name,
-        "description": request.description,
-        "websiteUrl": request.website_url,
-        "logoUrl": request.logo_url,
-        "metadata": prisma.Json(request.metadata or {}),
-    }
-    if creator_id:
-        record = await prisma.models.LlmModelCreator.prisma().update(
-            where={"id": creator_id},
-            data=data,
-        )
-    else:
-        record = await prisma.models.LlmModelCreator.prisma().create(data=data)
-    if record is None:
-        raise ValueError("Failed to create/update creator")
-    return _map_creator(record)
-
-
-async def delete_creator(creator_id: str) -> bool:
-    """
-    Delete a model creator.
-
-    This will set creatorId to NULL on all associated models (due to onDelete: SetNull).
-
-    Args:
-        creator_id: UUID of the creator to delete
-
-    Returns:
-        True if deleted successfully
-
-    Raises:
-        ValueError: If creator not found
-    """
-    creator = await prisma.models.LlmModelCreator.prisma().find_unique(
-        where={"id": creator_id}
-    )
-    if not creator:
-        raise ValueError(f"Creator with id '{creator_id}' not found")
-
-    await prisma.models.LlmModelCreator.prisma().delete(where={"id": creator_id})
-    return True
-
-
-async def get_recommended_model() -> llm_model.LlmModel | None:
-    """
-    Get the currently recommended LLM model.
-
-    Returns:
-        The recommended model, or None if no model is marked as recommended.
-    """
-    record = await prisma.models.LlmModel.prisma().find_first(
-        where={"isRecommended": True, "isEnabled": True},
-        include={"Costs": True, "Creator": True},
-    )
-    return _map_model(record) if record else None
-
-
-async def set_recommended_model(
-    model_id: str,
-) -> tuple[llm_model.LlmModel, str | None]:
-    """
-    Set a model as the recommended model.
-
-    This will clear the isRecommended flag from any other model and set it
-    on the specified model. The model must be enabled.
-
-    Args:
-        model_id: UUID of the model to set as recommended
-
-    Returns:
-        Tuple of (the updated model, previous recommended model slug or None)
-
-    Raises:
-        ValueError: If model not found or not enabled
-    """
-    # First, verify the model exists and is enabled
-    target_model = await prisma.models.LlmModel.prisma().find_unique(
-        where={"id": model_id}
-    )
-    if not target_model:
-        raise ValueError(f"Model with id '{model_id}' not found")
-    if not target_model.isEnabled:
-        raise ValueError(
-            f"Cannot set disabled model '{target_model.slug}' as recommended"
-        )
-
-    # Get the current recommended model (if any)
-    current_recommended = await prisma.models.LlmModel.prisma().find_first(
-        where={"isRecommended": True}
-    )
-    previous_slug = current_recommended.slug if current_recommended else None
-
-    # Use a transaction to ensure atomicity
-    async with transaction() as tx:
-        # Clear isRecommended from all models
-        await tx.llmmodel.update_many(
-            where={"isRecommended": True},
-            data={"isRecommended": False},
-        )
-        # Set the new recommended model
-        await tx.llmmodel.update(
-            where={"id": model_id},
-            data={"isRecommended": True},
-        )
-
-    # Fetch and return the updated model
-    updated_record = await prisma.models.LlmModel.prisma().find_unique(
-        where={"id": model_id},
-        include={"Costs": True, "Creator": True},
-    )
-    if not updated_record:
-        raise ValueError("Failed to fetch updated model")
-
-    return _map_model(updated_record), previous_slug
-
-
-async def get_recommended_model_slug() -> str | None:
-    """
-    Get the slug of the currently recommended LLM model.
-
-    Returns:
-        The slug of the recommended model, or None if no model is marked as recommended.
-    """
-    record = await prisma.models.LlmModel.prisma().find_first(
-        where={"isRecommended": True, "isEnabled": True},
-    )
-    return record.slug if record else None

autogpt_platform/backend/backend/server/v2/llm/model.py

@@ -1,235 +0,0 @@
-from __future__ import annotations
-
-import re
-from datetime import datetime
-from typing import Any, Optional
-
-import prisma.enums
-import pydantic
-
-from backend.util.models import Pagination
-
-# Pattern for valid model slugs: alphanumeric start, then alphanumeric, dots, underscores, slashes, hyphens
-SLUG_PATTERN = re.compile(r"^[a-zA-Z0-9][a-zA-Z0-9._/-]*$")
-
-
-class LlmModelCost(pydantic.BaseModel):
-    id: str
-    unit: prisma.enums.LlmCostUnit = prisma.enums.LlmCostUnit.RUN
-    credit_cost: int
-    credential_provider: str
-    credential_id: Optional[str] = None
-    credential_type: Optional[str] = None
-    currency: Optional[str] = None
-    metadata: dict[str, Any] = pydantic.Field(default_factory=dict)
-
-
-class LlmModelCreator(pydantic.BaseModel):
-    """Represents the organization that created/trained the model (e.g., OpenAI, Meta)."""
-
-    id: str
-    name: str
-    display_name: str
-    description: Optional[str] = None
-    website_url: Optional[str] = None
-    logo_url: Optional[str] = None
-    metadata: dict[str, Any] = pydantic.Field(default_factory=dict)
-
-
-class LlmModel(pydantic.BaseModel):
-    id: str
-    slug: str
-    display_name: str
-    description: Optional[str] = None
-    provider_id: str
-    creator_id: Optional[str] = None
-    creator: Optional[LlmModelCreator] = None
-    context_window: int
-    max_output_tokens: Optional[int] = None
-    is_enabled: bool = True
-    is_recommended: bool = False
-    capabilities: dict[str, Any] = pydantic.Field(default_factory=dict)
-    metadata: dict[str, Any] = pydantic.Field(default_factory=dict)
-    costs: list[LlmModelCost] = pydantic.Field(default_factory=list)
-
-
-class LlmProvider(pydantic.BaseModel):
-    id: str
-    name: str
-    display_name: str
-    description: Optional[str] = None
-    default_credential_provider: Optional[str] = None
-    default_credential_id: Optional[str] = None
-    default_credential_type: Optional[str] = None
-    supports_tools: bool = True
-    supports_json_output: bool = True
-    supports_reasoning: bool = False
-    supports_parallel_tool: bool = False
-    metadata: dict[str, Any] = pydantic.Field(default_factory=dict)
-    models: list[LlmModel] = pydantic.Field(default_factory=list)
-
-
-class LlmProvidersResponse(pydantic.BaseModel):
-    providers: list[LlmProvider]
-
-
-class LlmModelsResponse(pydantic.BaseModel):
-    models: list[LlmModel]
-    pagination: Optional[Pagination] = None
-
-
-class LlmCreatorsResponse(pydantic.BaseModel):
-    creators: list[LlmModelCreator]
-
-
-class UpsertLlmProviderRequest(pydantic.BaseModel):
-    name: str
-    display_name: str
-    description: Optional[str] = None
-    default_credential_provider: Optional[str] = None
-    default_credential_id: Optional[str] = None
-    default_credential_type: Optional[str] = "api_key"
-    supports_tools: bool = True
-    supports_json_output: bool = True
-    supports_reasoning: bool = False
-    supports_parallel_tool: bool = False
-    metadata: dict[str, Any] = pydantic.Field(default_factory=dict)
-
-
-class UpsertLlmCreatorRequest(pydantic.BaseModel):
-    name: str
-    display_name: str
-    description: Optional[str] = None
-    website_url: Optional[str] = None
-    logo_url: Optional[str] = None
-    metadata: dict[str, Any] = pydantic.Field(default_factory=dict)
-
-
-class LlmModelCostInput(pydantic.BaseModel):
-    unit: prisma.enums.LlmCostUnit = prisma.enums.LlmCostUnit.RUN
-    credit_cost: int
-    credential_provider: str
-    credential_id: Optional[str] = None
-    credential_type: Optional[str] = "api_key"
-    currency: Optional[str] = None
-    metadata: dict[str, Any] = pydantic.Field(default_factory=dict)
-
-
-class CreateLlmModelRequest(pydantic.BaseModel):
-    slug: str
-    display_name: str
-    description: Optional[str] = None
-    provider_id: str
-    creator_id: Optional[str] = None
-    context_window: int
-    max_output_tokens: Optional[int] = None
-    is_enabled: bool = True
-    capabilities: dict[str, Any] = pydantic.Field(default_factory=dict)
-    metadata: dict[str, Any] = pydantic.Field(default_factory=dict)
-    costs: list[LlmModelCostInput]
-
-    @pydantic.field_validator("slug")
-    @classmethod
-    def validate_slug(cls, v: str) -> str:
-        if not v or len(v) > 100:
-            raise ValueError("Slug must be 1-100 characters")
-        if not SLUG_PATTERN.match(v):
-            raise ValueError(
-                "Slug must start with alphanumeric and contain only "
-                "alphanumeric characters, dots, underscores, slashes, or hyphens"
-            )
-        return v
-
-
-class UpdateLlmModelRequest(pydantic.BaseModel):
-    display_name: Optional[str] = None
-    description: Optional[str] = None
-    context_window: Optional[int] = None
-    max_output_tokens: Optional[int] = None
-    is_enabled: Optional[bool] = None
-    capabilities: Optional[dict[str, Any]] = None
-    metadata: Optional[dict[str, Any]] = None
-    provider_id: Optional[str] = None
-    creator_id: Optional[str] = None
-    costs: Optional[list[LlmModelCostInput]] = None
-
-
-class ToggleLlmModelRequest(pydantic.BaseModel):
-    is_enabled: bool
-    migrate_to_slug: Optional[str] = None
-    migration_reason: Optional[str] = None  # e.g., "Provider outage"
-    # Custom pricing override for migrated workflows. When set, billing should use
-    # this cost instead of the target model's cost for affected nodes.
-    # See LlmModelMigration in schema.prisma for full documentation.
-    custom_credit_cost: Optional[int] = None
-
-
-class ToggleLlmModelResponse(pydantic.BaseModel):
-    model: LlmModel
-    nodes_migrated: int = 0
-    migrated_to_slug: Optional[str] = None
-    migration_id: Optional[str] = None  # ID of the migration record for revert
-
-
-class DeleteLlmModelResponse(pydantic.BaseModel):
-    deleted_model_slug: str
-    deleted_model_display_name: str
-    replacement_model_slug: Optional[str] = None
-    nodes_migrated: int
-    message: str
-
-
-class LlmModelUsageResponse(pydantic.BaseModel):
-    model_slug: str
-    node_count: int
-
-
-# Migration tracking models
-class LlmModelMigration(pydantic.BaseModel):
-    id: str
-    source_model_slug: str
-    target_model_slug: str
-    reason: Optional[str] = None
-    node_count: int
-    # Custom pricing override - billing should use this instead of target model's cost
-    custom_credit_cost: Optional[int] = None
-    is_reverted: bool = False
-    created_at: datetime
-    reverted_at: Optional[datetime] = None
-
-
-class LlmMigrationsResponse(pydantic.BaseModel):
-    migrations: list[LlmModelMigration]
-
-
-class RevertMigrationRequest(pydantic.BaseModel):
-    re_enable_source_model: bool = (
-        True  # Whether to re-enable the source model if disabled
-    )
-
-
-class RevertMigrationResponse(pydantic.BaseModel):
-    migration_id: str
-    source_model_slug: str
-    target_model_slug: str
-    nodes_reverted: int
-    nodes_already_changed: int = (
-        0  # Nodes that were modified since migration (not reverted)
-    )
-    source_model_re_enabled: bool = False  # Whether the source model was re-enabled
-    message: str
-
-
-class SetRecommendedModelRequest(pydantic.BaseModel):
-    model_id: str
-
-
-class SetRecommendedModelResponse(pydantic.BaseModel):
-    model: LlmModel
-    previous_recommended_slug: Optional[str] = None
-    message: str
-
-
-class RecommendedModelResponse(pydantic.BaseModel):
-    model: Optional[LlmModel] = None
-    slug: Optional[str] = None

autogpt_platform/backend/backend/server/v2/llm/routes.py

@@ -1,29 +0,0 @@
-import autogpt_libs.auth
-import fastapi
-
-from backend.server.v2.llm import db as llm_db
-from backend.server.v2.llm import model as llm_model
-
-router = fastapi.APIRouter(
-    prefix="/llm",
-    tags=["llm"],
-    dependencies=[fastapi.Security(autogpt_libs.auth.requires_user)],
-)
-
-
-@router.get("/models", response_model=llm_model.LlmModelsResponse)
-async def list_models(
-    page: int = fastapi.Query(default=1, ge=1, description="Page number (1-indexed)"),
-    page_size: int = fastapi.Query(
-        default=50, ge=1, le=100, description="Number of models per page"
-    ),
-):
-    """List all enabled LLM models available to users."""
-    return await llm_db.list_models(enabled_only=True, page=page, page_size=page_size)
-
-
-@router.get("/providers", response_model=llm_model.LlmProvidersResponse)
-async def list_providers():
-    """List all LLM providers with their enabled models."""
-    providers = await llm_db.list_providers(include_models=True, enabled_only=True)
-    return llm_model.LlmProvidersResponse(providers=providers)

autogpt_platform/backend/backend/util/feature_flag.py

@@ -38,6 +38,7 @@ class Flag(str, Enum):
     AGENT_ACTIVITY = "agent-activity"
     ENABLE_PLATFORM_PAYMENT = "enable-platform-payment"
     CHAT = "chat"
+    COPILOT_SDK = "copilot-sdk"
 
 
 def is_configured() -> bool:

autogpt_platform/backend/backend/util/request.py

@@ -101,7 +101,7 @@ class HostResolver(abc.AbstractResolver):
     def __init__(self, ssl_hostname: str, ip_addresses: list[str]):
         self.ssl_hostname = ssl_hostname
         self.ip_addresses = ip_addresses
-        self._default = aiohttp.AsyncResolver()
+        self._default = aiohttp.ThreadedResolver()
 
     async def resolve(self, host, port=0, family=socket.AF_INET):
         if host == self.ssl_hostname:
@@ -467,7 +467,7 @@ class Requests:
             resolver = HostResolver(ssl_hostname=hostname, ip_addresses=ip_addresses)
             ssl_context = ssl.create_default_context()
             connector = aiohttp.TCPConnector(resolver=resolver, ssl=ssl_context)
-        session_kwargs = {}
+        session_kwargs: dict = {}
         if connector:
             session_kwargs["connector"] = connector
 

autogpt_platform/backend/backend/util/settings.py

@@ -662,6 +662,17 @@ class Secrets(UpdateTrackingModel["Secrets"], BaseSettings):
     mem0_api_key: str = Field(default="", description="Mem0 API key")
     elevenlabs_api_key: str = Field(default="", description="ElevenLabs API key")
 
+    linear_api_key: str = Field(
+        default="", description="Linear API key for system-level operations"
+    )
+    linear_feature_request_project_id: str = Field(
+        default="",
+        description="Linear project ID where feature requests are tracked",
+    )
+    linear_feature_request_team_id: str = Field(
+        default="",
+        description="Linear team ID used when creating feature request issues",
+    )
     linear_client_id: str = Field(default="", description="Linear client ID")
     linear_client_secret: str = Field(default="", description="Linear client secret")
 

autogpt_platform/backend/migrations/20251126113000_add_llm_registry/migration.sql

@@ -1,81 +0,0 @@
--- CreateEnum
-CREATE TYPE "LlmCostUnit" AS ENUM ('RUN', 'TOKENS');
-
--- CreateTable
-CREATE TABLE "LlmProvider" (
-    "id" TEXT NOT NULL,
-    "createdAt" TIMESTAMP(3) NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    "updatedAt" TIMESTAMP(3) NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    "name" TEXT NOT NULL,
-    "displayName" TEXT NOT NULL,
-    "description" TEXT,
-    "defaultCredentialProvider" TEXT,
-    "defaultCredentialId" TEXT,
-    "defaultCredentialType" TEXT,
-    "supportsTools" BOOLEAN NOT NULL DEFAULT TRUE,
-    "supportsJsonOutput" BOOLEAN NOT NULL DEFAULT TRUE,
-    "supportsReasoning" BOOLEAN NOT NULL DEFAULT FALSE,
-    "supportsParallelTool" BOOLEAN NOT NULL DEFAULT FALSE,
-    "metadata" JSONB NOT NULL DEFAULT '{}'::jsonb,
-
-    CONSTRAINT "LlmProvider_pkey" PRIMARY KEY ("id"),
-    CONSTRAINT "LlmProvider_name_key" UNIQUE ("name")
-);
-
--- CreateTable
-CREATE TABLE "LlmModel" (
-    "id" TEXT NOT NULL,
-    "createdAt" TIMESTAMP(3) NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    "updatedAt" TIMESTAMP(3) NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    "slug" TEXT NOT NULL,
-    "displayName" TEXT NOT NULL,
-    "description" TEXT,
-    "providerId" TEXT NOT NULL,
-    "contextWindow" INTEGER NOT NULL,
-    "maxOutputTokens" INTEGER,
-    "isEnabled" BOOLEAN NOT NULL DEFAULT TRUE,
-    "capabilities" JSONB NOT NULL DEFAULT '{}'::jsonb,
-    "metadata" JSONB NOT NULL DEFAULT '{}'::jsonb,
-
-    CONSTRAINT "LlmModel_pkey" PRIMARY KEY ("id"),
-    CONSTRAINT "LlmModel_slug_key" UNIQUE ("slug")
-);
-
--- CreateTable
-CREATE TABLE "LlmModelCost" (
-    "id" TEXT NOT NULL,
-    "createdAt" TIMESTAMP(3) NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    "updatedAt" TIMESTAMP(3) NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    "unit" "LlmCostUnit" NOT NULL DEFAULT 'RUN',
-    "creditCost" INTEGER NOT NULL,
-    "credentialProvider" TEXT NOT NULL,
-    "credentialId" TEXT,
-    "credentialType" TEXT,
-    "currency" TEXT,
-    "metadata" JSONB NOT NULL DEFAULT '{}'::jsonb,
-    "llmModelId" TEXT NOT NULL,
-
-    CONSTRAINT "LlmModelCost_pkey" PRIMARY KEY ("id")
-);
-
--- CreateIndex
-CREATE INDEX "LlmModel_providerId_isEnabled_idx" ON "LlmModel"("providerId", "isEnabled");
-
--- CreateIndex
-CREATE INDEX "LlmModel_slug_idx" ON "LlmModel"("slug");
-
--- CreateIndex
-CREATE INDEX "LlmModelCost_llmModelId_idx" ON "LlmModelCost"("llmModelId");
-
--- CreateIndex
-CREATE INDEX "LlmModelCost_credentialProvider_idx" ON "LlmModelCost"("credentialProvider");
-
--- CreateIndex
-CREATE UNIQUE INDEX "LlmModelCost_llmModelId_credentialProvider_unit_key" ON "LlmModelCost"("llmModelId", "credentialProvider", "unit");
-
--- AddForeignKey
-ALTER TABLE "LlmModel" ADD CONSTRAINT "LlmModel_providerId_fkey" FOREIGN KEY ("providerId") REFERENCES "LlmProvider"("id") ON DELETE RESTRICT ON UPDATE CASCADE;
-
--- AddForeignKey
-ALTER TABLE "LlmModelCost" ADD CONSTRAINT "LlmModelCost_llmModelId_fkey" FOREIGN KEY ("llmModelId") REFERENCES "LlmModel"("id") ON DELETE CASCADE ON UPDATE CASCADE;
-

autogpt_platform/backend/migrations/20251126120000_seed_llm_registry/migration.sql

@@ -1,226 +0,0 @@
--- Seed LLM Registry from existing hard-coded data
--- This migration populates the LlmProvider, LlmModel, and LlmModelCost tables
--- with data from the existing MODEL_METADATA and MODEL_COST dictionaries
-
--- Insert Providers
-INSERT INTO "LlmProvider" ("id", "name", "displayName", "description", "defaultCredentialProvider", "defaultCredentialType", "supportsTools", "supportsJsonOutput", "supportsReasoning", "supportsParallelTool", "metadata")
-VALUES
-    (gen_random_uuid(), 'openai', 'OpenAI', 'OpenAI language models', 'openai', 'api_key', true, true, true, true, '{}'::jsonb),
-    (gen_random_uuid(), 'anthropic', 'Anthropic', 'Anthropic Claude models', 'anthropic', 'api_key', true, true, true, false, '{}'::jsonb),
-    (gen_random_uuid(), 'groq', 'Groq', 'Groq inference API', 'groq', 'api_key', false, true, false, false, '{}'::jsonb),
-    (gen_random_uuid(), 'open_router', 'OpenRouter', 'OpenRouter unified API', 'open_router', 'api_key', true, true, false, false, '{}'::jsonb),
-    (gen_random_uuid(), 'aiml_api', 'AI/ML API', 'AI/ML API models', 'aiml_api', 'api_key', false, true, false, false, '{}'::jsonb),
-    (gen_random_uuid(), 'ollama', 'Ollama', 'Ollama local models', 'ollama', 'api_key', false, true, false, false, '{}'::jsonb),
-    (gen_random_uuid(), 'llama_api', 'Llama API', 'Llama API models', 'llama_api', 'api_key', false, true, false, false, '{}'::jsonb),
-    (gen_random_uuid(), 'v0', 'v0', 'v0 by Vercel models', 'v0', 'api_key', true, true, false, false, '{}'::jsonb)
-ON CONFLICT ("name") DO NOTHING;
-
--- Insert Models (using CTEs to reference provider IDs)
-WITH provider_ids AS (
-    SELECT "id", "name" FROM "LlmProvider"
-)
-INSERT INTO "LlmModel" ("id", "slug", "displayName", "description", "providerId", "contextWindow", "maxOutputTokens", "isEnabled", "capabilities", "metadata")
-SELECT
-    gen_random_uuid(),
-    model_slug,
-    model_display_name,
-    NULL,
-    p."id",
-    context_window,
-    max_output_tokens,
-    true,
-    '{}'::jsonb,
-    '{}'::jsonb
-FROM (VALUES
-    -- OpenAI models
-    ('o3', 'O3', 'openai', 200000, 100000),
-    ('o3-mini', 'O3 Mini', 'openai', 200000, 100000),
-    ('o1', 'O1', 'openai', 200000, 100000),
-    ('o1-mini', 'O1 Mini', 'openai', 128000, 65536),
-    ('gpt-5-2025-08-07', 'GPT 5', 'openai', 400000, 128000),
-    ('gpt-5.1-2025-11-13', 'GPT 5.1', 'openai', 400000, 128000),
-    ('gpt-5-mini-2025-08-07', 'GPT 5 Mini', 'openai', 400000, 128000),
-    ('gpt-5-nano-2025-08-07', 'GPT 5 Nano', 'openai', 400000, 128000),
-    ('gpt-5-chat-latest', 'GPT 5 Chat', 'openai', 400000, 16384),
-    ('gpt-4.1-2025-04-14', 'GPT 4.1', 'openai', 1000000, 32768),
-    ('gpt-4.1-mini-2025-04-14', 'GPT 4.1 Mini', 'openai', 1047576, 32768),
-    ('gpt-4o-mini', 'GPT 4o Mini', 'openai', 128000, 16384),
-    ('gpt-4o', 'GPT 4o', 'openai', 128000, 16384),
-    ('gpt-4-turbo', 'GPT 4 Turbo', 'openai', 128000, 4096),
-    ('gpt-3.5-turbo', 'GPT 3.5 Turbo', 'openai', 16385, 4096),
-    -- Anthropic models
-    ('claude-opus-4-1-20250805', 'Claude 4.1 Opus', 'anthropic', 200000, 32000),
-    ('claude-opus-4-20250514', 'Claude 4 Opus', 'anthropic', 200000, 32000),
-    ('claude-sonnet-4-20250514', 'Claude 4 Sonnet', 'anthropic', 200000, 64000),
-    ('claude-opus-4-5-20251101', 'Claude 4.5 Opus', 'anthropic', 200000, 64000),
-    ('claude-sonnet-4-5-20250929', 'Claude 4.5 Sonnet', 'anthropic', 200000, 64000),
-    ('claude-haiku-4-5-20251001', 'Claude 4.5 Haiku', 'anthropic', 200000, 64000),
-    ('claude-3-7-sonnet-20250219', 'Claude 3.7 Sonnet', 'anthropic', 200000, 64000),
-    ('claude-3-haiku-20240307', 'Claude 3 Haiku', 'anthropic', 200000, 4096),
-    -- AI/ML API models
-    ('Qwen/Qwen2.5-72B-Instruct-Turbo', 'Qwen 2.5 72B', 'aiml_api', 32000, 8000),
-    ('nvidia/llama-3.1-nemotron-70b-instruct', 'Llama 3.1 Nemotron 70B', 'aiml_api', 128000, 40000),
-    ('meta-llama/Llama-3.3-70B-Instruct-Turbo', 'Llama 3.3 70B', 'aiml_api', 128000, NULL),
-    ('meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo', 'Meta Llama 3.1 70B', 'aiml_api', 131000, 2000),
-    ('meta-llama/Llama-3.2-3B-Instruct-Turbo', 'Llama 3.2 3B', 'aiml_api', 128000, NULL),
-    -- Groq models
-    ('llama-3.3-70b-versatile', 'Llama 3.3 70B', 'groq', 128000, 32768),
-    ('llama-3.1-8b-instant', 'Llama 3.1 8B', 'groq', 128000, 8192),
-    -- Ollama models
-    ('llama3.3', 'Llama 3.3', 'ollama', 8192, NULL),
-    ('llama3.2', 'Llama 3.2', 'ollama', 8192, NULL),
-    ('llama3', 'Llama 3', 'ollama', 8192, NULL),
-    ('llama3.1:405b', 'Llama 3.1 405B', 'ollama', 8192, NULL),
-    ('dolphin-mistral:latest', 'Dolphin Mistral', 'ollama', 32768, NULL),
-    -- OpenRouter models
-    ('google/gemini-2.5-pro-preview-03-25', 'Gemini 2.5 Pro', 'open_router', 1050000, 8192),
-    ('google/gemini-3-pro-preview', 'Gemini 3 Pro Preview', 'open_router', 1048576, 65535),
-    ('google/gemini-2.5-flash', 'Gemini 2.5 Flash', 'open_router', 1048576, 65535),
-    ('google/gemini-2.0-flash-001', 'Gemini 2.0 Flash', 'open_router', 1048576, 8192),
-    ('google/gemini-2.5-flash-lite-preview-06-17', 'Gemini 2.5 Flash Lite Preview', 'open_router', 1048576, 65535),
-    ('google/gemini-2.0-flash-lite-001', 'Gemini 2.0 Flash Lite', 'open_router', 1048576, 8192),
-    ('mistralai/mistral-nemo', 'Mistral Nemo', 'open_router', 128000, 4096),
-    ('cohere/command-r-08-2024', 'Command R', 'open_router', 128000, 4096),
-    ('cohere/command-r-plus-08-2024', 'Command R Plus', 'open_router', 128000, 4096),
-    ('deepseek/deepseek-chat', 'DeepSeek Chat', 'open_router', 64000, 2048),
-    ('deepseek/deepseek-r1-0528', 'DeepSeek R1', 'open_router', 163840, 163840),
-    ('perplexity/sonar', 'Perplexity Sonar', 'open_router', 127000, 8000),
-    ('perplexity/sonar-pro', 'Perplexity Sonar Pro', 'open_router', 200000, 8000),
-    ('perplexity/sonar-deep-research', 'Perplexity Sonar Deep Research', 'open_router', 128000, 16000),
-    ('nousresearch/hermes-3-llama-3.1-405b', 'Hermes 3 Llama 3.1 405B', 'open_router', 131000, 4096),
-    ('nousresearch/hermes-3-llama-3.1-70b', 'Hermes 3 Llama 3.1 70B', 'open_router', 12288, 12288),
-    ('openai/gpt-oss-120b', 'GPT OSS 120B', 'open_router', 131072, 131072),
-    ('openai/gpt-oss-20b', 'GPT OSS 20B', 'open_router', 131072, 32768),
-    ('amazon/nova-lite-v1', 'Amazon Nova Lite', 'open_router', 300000, 5120),
-    ('amazon/nova-micro-v1', 'Amazon Nova Micro', 'open_router', 128000, 5120),
-    ('amazon/nova-pro-v1', 'Amazon Nova Pro', 'open_router', 300000, 5120),
-    ('microsoft/wizardlm-2-8x22b', 'WizardLM 2 8x22B', 'open_router', 65536, 4096),
-    ('gryphe/mythomax-l2-13b', 'MythoMax L2 13B', 'open_router', 4096, 4096),
-    ('meta-llama/llama-4-scout', 'Llama 4 Scout', 'open_router', 131072, 131072),
-    ('meta-llama/llama-4-maverick', 'Llama 4 Maverick', 'open_router', 1048576, 1000000),
-    ('x-ai/grok-4', 'Grok 4', 'open_router', 256000, 256000),
-    ('x-ai/grok-4-fast', 'Grok 4 Fast', 'open_router', 2000000, 30000),
-    ('x-ai/grok-4.1-fast', 'Grok 4.1 Fast', 'open_router', 2000000, 30000),
-    ('x-ai/grok-code-fast-1', 'Grok Code Fast 1', 'open_router', 256000, 10000),
-    ('moonshotai/kimi-k2', 'Kimi K2', 'open_router', 131000, 131000),
-    ('qwen/qwen3-235b-a22b-thinking-2507', 'Qwen 3 235B Thinking', 'open_router', 262144, 262144),
-    ('qwen/qwen3-coder', 'Qwen 3 Coder', 'open_router', 262144, 262144),
-    -- Llama API models
-    ('Llama-4-Scout-17B-16E-Instruct-FP8', 'Llama 4 Scout', 'llama_api', 128000, 4028),
-    ('Llama-4-Maverick-17B-128E-Instruct-FP8', 'Llama 4 Maverick', 'llama_api', 128000, 4028),
-    ('Llama-3.3-8B-Instruct', 'Llama 3.3 8B', 'llama_api', 128000, 4028),
-    ('Llama-3.3-70B-Instruct', 'Llama 3.3 70B', 'llama_api', 128000, 4028),
-    -- v0 models
-    ('v0-1.5-md', 'v0 1.5 MD', 'v0', 128000, 64000),
-    ('v0-1.5-lg', 'v0 1.5 LG', 'v0', 512000, 64000),
-    ('v0-1.0-md', 'v0 1.0 MD', 'v0', 128000, 64000)
-) AS models(model_slug, model_display_name, provider_name, context_window, max_output_tokens)
-JOIN provider_ids p ON p."name" = models.provider_name
-ON CONFLICT ("slug") DO NOTHING;
-
--- Insert Costs (using CTEs to reference model IDs)
-WITH model_ids AS (
-    SELECT "id", "slug", "providerId" FROM "LlmModel"
-),
-provider_ids AS (
-    SELECT "id", "name" FROM "LlmProvider"
-)
-INSERT INTO "LlmModelCost" ("id", "unit", "creditCost", "credentialProvider", "credentialId", "credentialType", "currency", "metadata", "llmModelId")
-SELECT
-    gen_random_uuid(),
-    'RUN'::"LlmCostUnit",
-    cost,
-    p."name",
-    NULL,
-    'api_key',
-    NULL,
-    '{}'::jsonb,
-    m."id"
-FROM (VALUES
-    -- OpenAI costs
-    ('o3', 4),
-    ('o3-mini', 2),
-    ('o1', 16),
-    ('o1-mini', 4),
-    ('gpt-5-2025-08-07', 2),
-    ('gpt-5.1-2025-11-13', 5),
-    ('gpt-5-mini-2025-08-07', 1),
-    ('gpt-5-nano-2025-08-07', 1),
-    ('gpt-5-chat-latest', 5),
-    ('gpt-4.1-2025-04-14', 2),
-    ('gpt-4.1-mini-2025-04-14', 1),
-    ('gpt-4o-mini', 1),
-    ('gpt-4o', 3),
-    ('gpt-4-turbo', 10),
-    ('gpt-3.5-turbo', 1),
-    -- Anthropic costs
-    ('claude-opus-4-1-20250805', 21),
-    ('claude-opus-4-20250514', 21),
-    ('claude-sonnet-4-20250514', 5),
-    ('claude-haiku-4-5-20251001', 4),
-    ('claude-opus-4-5-20251101', 14),
-    ('claude-sonnet-4-5-20250929', 9),
-    ('claude-3-7-sonnet-20250219', 5),
-    ('claude-3-haiku-20240307', 1),
-    -- AI/ML API costs
-    ('Qwen/Qwen2.5-72B-Instruct-Turbo', 1),
-    ('nvidia/llama-3.1-nemotron-70b-instruct', 1),
-    ('meta-llama/Llama-3.3-70B-Instruct-Turbo', 1),
-    ('meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo', 1),
-    ('meta-llama/Llama-3.2-3B-Instruct-Turbo', 1),
-    -- Groq costs
-    ('llama-3.3-70b-versatile', 1),
-    ('llama-3.1-8b-instant', 1),
-    -- Ollama costs
-    ('llama3.3', 1),
-    ('llama3.2', 1),
-    ('llama3', 1),
-    ('llama3.1:405b', 1),
-    ('dolphin-mistral:latest', 1),
-    -- OpenRouter costs
-    ('google/gemini-2.5-pro-preview-03-25', 4),
-    ('google/gemini-3-pro-preview', 5),
-    ('mistralai/mistral-nemo', 1),
-    ('cohere/command-r-08-2024', 1),
-    ('cohere/command-r-plus-08-2024', 3),
-    ('deepseek/deepseek-chat', 2),
-    ('perplexity/sonar', 1),
-    ('perplexity/sonar-pro', 5),
-    ('perplexity/sonar-deep-research', 10),
-    ('nousresearch/hermes-3-llama-3.1-405b', 1),
-    ('nousresearch/hermes-3-llama-3.1-70b', 1),
-    ('amazon/nova-lite-v1', 1),
-    ('amazon/nova-micro-v1', 1),
-    ('amazon/nova-pro-v1', 1),
-    ('microsoft/wizardlm-2-8x22b', 1),
-    ('gryphe/mythomax-l2-13b', 1),
-    ('meta-llama/llama-4-scout', 1),
-    ('meta-llama/llama-4-maverick', 1),
-    ('x-ai/grok-4', 9),
-    ('x-ai/grok-4-fast', 1),
-    ('x-ai/grok-4.1-fast', 1),
-    ('x-ai/grok-code-fast-1', 1),
-    ('moonshotai/kimi-k2', 1),
-    ('qwen/qwen3-235b-a22b-thinking-2507', 1),
-    ('qwen/qwen3-coder', 9),
-    ('google/gemini-2.5-flash', 1),
-    ('google/gemini-2.0-flash-001', 1),
-    ('google/gemini-2.5-flash-lite-preview-06-17', 1),
-    ('google/gemini-2.0-flash-lite-001', 1),
-    ('deepseek/deepseek-r1-0528', 1),
-    ('openai/gpt-oss-120b', 1),
-    ('openai/gpt-oss-20b', 1),
-    -- Llama API costs
-    ('Llama-4-Scout-17B-16E-Instruct-FP8', 1),
-    ('Llama-4-Maverick-17B-128E-Instruct-FP8', 1),
-    ('Llama-3.3-8B-Instruct', 1),
-    ('Llama-3.3-70B-Instruct', 1),
-    -- v0 costs
-    ('v0-1.5-md', 1),
-    ('v0-1.5-lg', 2),
-    ('v0-1.0-md', 1)
-) AS costs(model_slug, cost)
-JOIN model_ids m ON m."slug" = costs.model_slug
-JOIN provider_ids p ON p."id" = m."providerId"
-ON CONFLICT ("llmModelId", "credentialProvider", "unit") DO NOTHING;
-

autogpt_platform/backend/migrations/20251218100000_add_llm_model_migration/migration.sql

@@ -1,25 +0,0 @@
--- CreateTable
-CREATE TABLE "LlmModelMigration" (
-    "id" TEXT NOT NULL,
-    "createdAt" TIMESTAMP(3) NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    "updatedAt" TIMESTAMP(3) NOT NULL,
-    "sourceModelSlug" TEXT NOT NULL,
-    "targetModelSlug" TEXT NOT NULL,
-    "reason" TEXT,
-    "migratedNodeIds" JSONB NOT NULL DEFAULT '[]',
-    "nodeCount" INTEGER NOT NULL,
-    "customCreditCost" INTEGER,
-    "isReverted" BOOLEAN NOT NULL DEFAULT false,
-    "revertedAt" TIMESTAMP(3),
-
-    CONSTRAINT "LlmModelMigration_pkey" PRIMARY KEY ("id")
-);
-
--- CreateIndex
-CREATE INDEX "LlmModelMigration_sourceModelSlug_idx" ON "LlmModelMigration"("sourceModelSlug");
-
--- CreateIndex
-CREATE INDEX "LlmModelMigration_targetModelSlug_idx" ON "LlmModelMigration"("targetModelSlug");
-
--- CreateIndex
-CREATE INDEX "LlmModelMigration_isReverted_idx" ON "LlmModelMigration"("isReverted");

autogpt_platform/backend/migrations/20251224100000_add_llm_model_creator/migration.sql

@@ -1,127 +0,0 @@
--- Add LlmModelCreator table
--- Creator represents who made/trained the model (e.g., OpenAI, Meta)
--- This is distinct from Provider who hosts/serves the model (e.g., OpenRouter)
-
--- Create the LlmModelCreator table
-CREATE TABLE "LlmModelCreator" (
-    "id" TEXT NOT NULL,
-    "createdAt" TIMESTAMP(3) NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    "updatedAt" TIMESTAMP(3) NOT NULL,
-    "name" TEXT NOT NULL,
-    "displayName" TEXT NOT NULL,
-    "description" TEXT,
-    "websiteUrl" TEXT,
-    "logoUrl" TEXT,
-    "metadata" JSONB NOT NULL DEFAULT '{}',
-
-    CONSTRAINT "LlmModelCreator_pkey" PRIMARY KEY ("id")
-);
-
--- Create unique index on name
-CREATE UNIQUE INDEX "LlmModelCreator_name_key" ON "LlmModelCreator"("name");
-
--- Add creatorId column to LlmModel
-ALTER TABLE "LlmModel" ADD COLUMN "creatorId" TEXT;
-
--- Add foreign key constraint
-ALTER TABLE "LlmModel" ADD CONSTRAINT "LlmModel_creatorId_fkey"
-    FOREIGN KEY ("creatorId") REFERENCES "LlmModelCreator"("id") ON DELETE SET NULL ON UPDATE CASCADE;
-
--- Create index on creatorId
-CREATE INDEX "LlmModel_creatorId_idx" ON "LlmModel"("creatorId");
-
--- Seed creators based on known model creators
-INSERT INTO "LlmModelCreator" ("id", "updatedAt", "name", "displayName", "description", "websiteUrl", "metadata")
-VALUES
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'openai', 'OpenAI', 'Creator of GPT models', 'https://openai.com', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'anthropic', 'Anthropic', 'Creator of Claude models', 'https://anthropic.com', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'meta', 'Meta', 'Creator of Llama models', 'https://ai.meta.com', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'google', 'Google', 'Creator of Gemini models', 'https://deepmind.google', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'mistral', 'Mistral AI', 'Creator of Mistral models', 'https://mistral.ai', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'cohere', 'Cohere', 'Creator of Command models', 'https://cohere.com', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'deepseek', 'DeepSeek', 'Creator of DeepSeek models', 'https://deepseek.com', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'perplexity', 'Perplexity AI', 'Creator of Sonar models', 'https://perplexity.ai', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'qwen', 'Qwen (Alibaba)', 'Creator of Qwen models', 'https://qwenlm.github.io', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'xai', 'xAI', 'Creator of Grok models', 'https://x.ai', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'amazon', 'Amazon', 'Creator of Nova models', 'https://aws.amazon.com/bedrock', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'microsoft', 'Microsoft', 'Creator of WizardLM models', 'https://microsoft.com', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'moonshot', 'Moonshot AI', 'Creator of Kimi models', 'https://moonshot.cn', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'nvidia', 'NVIDIA', 'Creator of Nemotron models', 'https://nvidia.com', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'nous_research', 'Nous Research', 'Creator of Hermes models', 'https://nousresearch.com', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'vercel', 'Vercel', 'Creator of v0 models', 'https://vercel.com', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'cognitive_computations', 'Cognitive Computations', 'Creator of Dolphin models', 'https://erichartford.com', '{}'),
-    (gen_random_uuid(), CURRENT_TIMESTAMP, 'gryphe', 'Gryphe', 'Creator of MythoMax models', 'https://huggingface.co/Gryphe', '{}')
-ON CONFLICT ("name") DO NOTHING;
-
--- Update existing models with their creators
--- OpenAI models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'openai')
-WHERE "slug" LIKE 'gpt-%' OR "slug" LIKE 'o1%' OR "slug" LIKE 'o3%' OR "slug" LIKE 'openai/%';
-
--- Anthropic models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'anthropic')
-WHERE "slug" LIKE 'claude-%';
-
--- Meta/Llama models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'meta')
-WHERE "slug" LIKE 'llama%' OR "slug" LIKE 'Llama%' OR "slug" LIKE 'meta-llama/%' OR "slug" LIKE '%/llama-%';
-
--- Google models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'google')
-WHERE "slug" LIKE 'google/%' OR "slug" LIKE 'gemini%';
-
--- Mistral models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'mistral')
-WHERE "slug" LIKE 'mistral%' OR "slug" LIKE 'mistralai/%';
-
--- Cohere models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'cohere')
-WHERE "slug" LIKE 'cohere/%' OR "slug" LIKE 'command-%';
-
--- DeepSeek models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'deepseek')
-WHERE "slug" LIKE 'deepseek/%' OR "slug" LIKE 'deepseek-%';
-
--- Perplexity models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'perplexity')
-WHERE "slug" LIKE 'perplexity/%' OR "slug" LIKE 'sonar%';
-
--- Qwen models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'qwen')
-WHERE "slug" LIKE 'Qwen/%' OR "slug" LIKE 'qwen/%';
-
--- xAI/Grok models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'xai')
-WHERE "slug" LIKE 'x-ai/%' OR "slug" LIKE 'grok%';
-
--- Amazon models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'amazon')
-WHERE "slug" LIKE 'amazon/%' OR "slug" LIKE 'nova-%';
-
--- Microsoft models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'microsoft')
-WHERE "slug" LIKE 'microsoft/%' OR "slug" LIKE 'wizardlm%';
-
--- Moonshot models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'moonshot')
-WHERE "slug" LIKE 'moonshotai/%' OR "slug" LIKE 'kimi%';
-
--- NVIDIA models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'nvidia')
-WHERE "slug" LIKE 'nvidia/%' OR "slug" LIKE '%nemotron%';
-
--- Nous Research models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'nous_research')
-WHERE "slug" LIKE 'nousresearch/%' OR "slug" LIKE 'hermes%';
-
--- Vercel/v0 models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'vercel')
-WHERE "slug" LIKE 'v0-%';
-
--- Dolphin models (Cognitive Computations / Eric Hartford)
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'cognitive_computations')
-WHERE "slug" LIKE 'dolphin-%';
-
--- Gryphe models
-UPDATE "LlmModel" SET "creatorId" = (SELECT "id" FROM "LlmModelCreator" WHERE "name" = 'gryphe')
-WHERE "slug" LIKE 'gryphe/%' OR "slug" LIKE 'mythomax%';

autogpt_platform/backend/migrations/20260105120000_add_agent_node_model_index/migration.sql

@@ -1,4 +0,0 @@
--- CreateIndex
--- Index for efficient LLM model lookups on AgentNode.constantInput->>'model'
--- This improves performance of model migration queries in the LLM registry
-CREATE INDEX "AgentNode_constantInput_model_idx" ON "AgentNode" ((("constantInput"->>'model')));

autogpt_platform/backend/migrations/20260106150000_add_gpt52_model/migration.sql

@@ -1,52 +0,0 @@
--- Add GPT-5.2 model and update O3 slug
--- This migration adds the new GPT-5.2 model added in dev branch
-
--- Update O3 slug to match dev branch format
-UPDATE "LlmModel"
-SET "slug" = 'o3-2025-04-16'
-WHERE "slug" = 'o3';
-
--- Update cost reference for O3 if needed
--- (costs are linked by model ID, so no update needed)
-
--- Add GPT-5.2 model
-WITH provider_id AS (
-    SELECT "id" FROM "LlmProvider" WHERE "name" = 'openai'
-)
-INSERT INTO "LlmModel" ("id", "slug", "displayName", "description", "providerId", "contextWindow", "maxOutputTokens", "isEnabled", "capabilities", "metadata")
-SELECT
-    gen_random_uuid(),
-    'gpt-5.2-2025-12-11',
-    'GPT 5.2',
-    'OpenAI GPT-5.2 model',
-    p."id",
-    400000,
-    128000,
-    true,
-    '{}'::jsonb,
-    '{}'::jsonb
-FROM provider_id p
-ON CONFLICT ("slug") DO NOTHING;
-
--- Add cost for GPT-5.2
-WITH model_id AS (
-    SELECT m."id", p."name" as provider_name
-    FROM "LlmModel" m
-    JOIN "LlmProvider" p ON p."id" = m."providerId"
-    WHERE m."slug" = 'gpt-5.2-2025-12-11'
-)
-INSERT INTO "LlmModelCost" ("id", "unit", "creditCost", "credentialProvider", "credentialId", "credentialType", "currency", "metadata", "llmModelId")
-SELECT
-    gen_random_uuid(),
-    'RUN'::"LlmCostUnit",
-    3,  -- Same cost tier as GPT-5.1
-    m.provider_name,
-    NULL,
-    'api_key',
-    NULL,
-    '{}'::jsonb,
-    m."id"
-FROM model_id m
-WHERE NOT EXISTS (
-    SELECT 1 FROM "LlmModelCost" c WHERE c."llmModelId" = m."id"
-);

autogpt_platform/backend/migrations/20260107100000_add_llm_recommended_model/migration.sql

@@ -1,11 +0,0 @@
--- Add isRecommended field to LlmModel table
--- This allows admins to mark a model as the recommended default
-
-ALTER TABLE "LlmModel" ADD COLUMN "isRecommended" BOOLEAN NOT NULL DEFAULT false;
-
--- Set gpt-4o-mini as the default recommended model (if it exists)
-UPDATE "LlmModel" SET "isRecommended" = true WHERE "slug" = 'gpt-4o-mini' AND "isEnabled" = true;
-
--- Create unique partial index to enforce only one recommended model at the database level
--- This prevents multiple rows from having isRecommended = true
-CREATE UNIQUE INDEX "LlmModel_single_recommended_idx" ON "LlmModel" ("isRecommended") WHERE "isRecommended" = true;

autogpt_platform/backend/migrations/20260122000000_add_llm_price_tier/migration.sql

@@ -1,61 +0,0 @@
--- Add new columns to LlmModel table for extended model metadata
--- These columns support the LLM Picker UI enhancements
-
--- Add priceTier column: 1=cheapest, 2=medium, 3=expensive
-ALTER TABLE "LlmModel" ADD COLUMN IF NOT EXISTS "priceTier" INTEGER NOT NULL DEFAULT 1;
-
--- Add creatorId column for model creator relationship (if not exists)
-ALTER TABLE "LlmModel" ADD COLUMN IF NOT EXISTS "creatorId" TEXT;
-
--- Add isRecommended column (if not exists)
-ALTER TABLE "LlmModel" ADD COLUMN IF NOT EXISTS "isRecommended" BOOLEAN NOT NULL DEFAULT FALSE;
-
--- Add index on creatorId if not exists
-CREATE INDEX IF NOT EXISTS "LlmModel_creatorId_idx" ON "LlmModel"("creatorId");
-
--- Add foreign key for creatorId if not exists
-DO $$
-BEGIN
-    IF NOT EXISTS (SELECT 1 FROM pg_constraint WHERE conname = 'LlmModel_creatorId_fkey') THEN
-        -- Only add FK if LlmModelCreator table exists
-        IF EXISTS (SELECT 1 FROM information_schema.tables WHERE table_name = 'LlmModelCreator') THEN
-            ALTER TABLE "LlmModel" ADD CONSTRAINT "LlmModel_creatorId_fkey"
-            FOREIGN KEY ("creatorId") REFERENCES "LlmModelCreator"("id") ON DELETE SET NULL ON UPDATE CASCADE;
-        END IF;
-    END IF;
-END $$;
-
--- Update priceTier values for existing models based on original MODEL_METADATA
--- Tier 1 = cheapest, Tier 2 = medium, Tier 3 = expensive
-
--- OpenAI models
-UPDATE "LlmModel" SET "priceTier" = 2 WHERE "slug" = 'o3';
-UPDATE "LlmModel" SET "priceTier" = 1 WHERE "slug" = 'o3-mini';
-UPDATE "LlmModel" SET "priceTier" = 3 WHERE "slug" = 'o1';
-UPDATE "LlmModel" SET "priceTier" = 2 WHERE "slug" = 'o1-mini';
-UPDATE "LlmModel" SET "priceTier" = 3 WHERE "slug" = 'gpt-5.2';
-UPDATE "LlmModel" SET "priceTier" = 2 WHERE "slug" = 'gpt-5.1';
-UPDATE "LlmModel" SET "priceTier" = 1 WHERE "slug" = 'gpt-5';
-UPDATE "LlmModel" SET "priceTier" = 1 WHERE "slug" = 'gpt-5-mini';
-UPDATE "LlmModel" SET "priceTier" = 1 WHERE "slug" = 'gpt-5-nano';
-UPDATE "LlmModel" SET "priceTier" = 2 WHERE "slug" = 'gpt-5-chat-latest';
-UPDATE "LlmModel" SET "priceTier" = 1 WHERE "slug" LIKE 'gpt-4.1%';
-UPDATE "LlmModel" SET "priceTier" = 1 WHERE "slug" = 'gpt-4o-mini';
-UPDATE "LlmModel" SET "priceTier" = 2 WHERE "slug" = 'gpt-4o';
-UPDATE "LlmModel" SET "priceTier" = 3 WHERE "slug" = 'gpt-4-turbo';
-UPDATE "LlmModel" SET "priceTier" = 1 WHERE "slug" = 'gpt-3.5-turbo';
-
--- Anthropic models
-UPDATE "LlmModel" SET "priceTier" = 3 WHERE "slug" LIKE 'claude-opus%';
-UPDATE "LlmModel" SET "priceTier" = 2 WHERE "slug" LIKE 'claude-sonnet%';
-UPDATE "LlmModel" SET "priceTier" = 3 WHERE "slug" LIKE 'claude%-4-5-sonnet%';
-UPDATE "LlmModel" SET "priceTier" = 2 WHERE "slug" LIKE 'claude%-haiku%';
-UPDATE "LlmModel" SET "priceTier" = 1 WHERE "slug" = 'claude-3-haiku-20240307';
-
--- OpenRouter models - Pro/expensive tiers
-UPDATE "LlmModel" SET "priceTier" = 2 WHERE "slug" LIKE 'google/gemini%-pro%';
-UPDATE "LlmModel" SET "priceTier" = 2 WHERE "slug" LIKE '%command-r-plus%';
-UPDATE "LlmModel" SET "priceTier" = 2 WHERE "slug" LIKE '%sonar-pro%';
-UPDATE "LlmModel" SET "priceTier" = 3 WHERE "slug" LIKE '%sonar-deep-research%';
-UPDATE "LlmModel" SET "priceTier" = 3 WHERE "slug" = 'x-ai/grok-4';
-UPDATE "LlmModel" SET "priceTier" = 3 WHERE "slug" LIKE '%qwen3-coder%';

autogpt_platform/backend/migrations/20260212000000_add_llm_migration_composite_index/migration.sql

@@ -1,6 +0,0 @@
--- Add composite index on LlmModelMigration for optimized active migration queries
--- This index improves performance when querying for non-reverted migrations by model slug
--- Used by the billing system to apply customCreditCost overrides
-
--- CreateIndex
-CREATE INDEX "LlmModelMigration_sourceModelSlug_isReverted_idx" ON "LlmModelMigration"("sourceModelSlug", "isReverted");

autogpt_platform/backend/migrations/20260212000001_sync_llm_models_with_dev/migration.sql

@@ -1,65 +0,0 @@
--- Sync LLM models with latest dev branch changes
--- This migration adds new models and removes deprecated ones
-
--- Remove models that were deleted from dev
-DELETE FROM "LlmModelCost" WHERE "llmModelId" IN (
-    SELECT "id" FROM "LlmModel" WHERE "slug" IN ('o3', 'o3-mini', 'claude-3-7-sonnet-20250219')
-);
-
-DELETE FROM "LlmModel" WHERE "slug" IN ('o3', 'o3-mini', 'claude-3-7-sonnet-20250219');
-
--- Add new models from dev
-WITH provider_ids AS (
-    SELECT "id", "name" FROM "LlmProvider"
-)
-INSERT INTO "LlmModel" ("id", "slug", "displayName", "description", "providerId", "contextWindow", "maxOutputTokens", "isEnabled", "capabilities", "metadata", "createdAt", "updatedAt")
-SELECT
-    gen_random_uuid(),
-    model_slug,
-    model_display_name,
-    NULL,
-    p."id",
-    context_window,
-    max_output_tokens,
-    true,
-    '{}'::jsonb,
-    '{}'::jsonb,
-    NOW(),
-    NOW()
-FROM (VALUES
-    -- New OpenAI model
-    ('gpt-5.2-2025-12-11', 'GPT 5.2', 'openai', 400000, 128000),
-    -- New Anthropic model
-    ('claude-opus-4-6', 'Claude 4.6 Opus', 'anthropic', 200000, 64000)
-) AS models(model_slug, model_display_name, provider_name, context_window, max_output_tokens)
-JOIN provider_ids p ON p."name" = models.provider_name
-ON CONFLICT ("slug") DO NOTHING;
-
--- Add costs for new models
-WITH model_ids AS (
-    SELECT "id", "slug", "providerId" FROM "LlmModel"
-),
-provider_ids AS (
-    SELECT "id", "name" FROM "LlmProvider"
-)
-INSERT INTO "LlmModelCost" ("id", "unit", "creditCost", "credentialProvider", "credentialId", "credentialType", "currency", "metadata", "llmModelId", "createdAt", "updatedAt")
-SELECT
-    gen_random_uuid(),
-    'RUN'::"LlmCostUnit",
-    cost,
-    p."name",
-    NULL,
-    'api_key',
-    NULL,
-    '{}'::jsonb,
-    m."id",
-    NOW(),
-    NOW()
-FROM (VALUES
-    -- New model costs (estimate based on similar models)
-    ('gpt-5.2-2025-12-11', 5),  -- Similar to GPT 5.1
-    ('claude-opus-4-6', 21)     -- Similar to other Opus 4.x models
-) AS costs(model_slug, cost)
-JOIN model_ids m ON m."slug" = costs.model_slug
-JOIN provider_ids p ON p."id" = m."providerId"
-ON CONFLICT ("llmModelId", "credentialProvider", "unit") DO NOTHING;

autogpt_platform/backend/poetry.lock

@@ -897,6 +897,29 @@ files = [
     {file = "charset_normalizer-3.4.4.tar.gz", hash = "sha256:94537985111c35f28720e43603b8e7b43a6ecfb2ce1d3058bbe955b73404e21a"},
 ]
 
+[[package]]
+name = "claude-agent-sdk"
+version = "0.1.35"
+description = "Python SDK for Claude Code"
+optional = false
+python-versions = ">=3.10"
+groups = ["main"]
+files = [
+    {file = "claude_agent_sdk-0.1.35-py3-none-macosx_11_0_arm64.whl", hash = "sha256:df67f4deade77b16a9678b3a626c176498e40417f33b04beda9628287f375591"},
+    {file = "claude_agent_sdk-0.1.35-py3-none-manylinux_2_17_aarch64.whl", hash = "sha256:14963944f55ded7c8ed518feebfa5b4284aa6dd8d81aeff2e5b21a962ce65097"},
+    {file = "claude_agent_sdk-0.1.35-py3-none-manylinux_2_17_x86_64.whl", hash = "sha256:84344dcc535d179c1fc8a11c6f34c37c3b583447bdf09d869effb26514fd7a65"},
+    {file = "claude_agent_sdk-0.1.35-py3-none-win_amd64.whl", hash = "sha256:1b3d54b47448c93f6f372acd4d1757f047c3c1e8ef5804be7a1e3e53e2c79a5f"},
+    {file = "claude_agent_sdk-0.1.35.tar.gz", hash = "sha256:0f98e2b3c71ca85abfc042e7a35c648df88e87fda41c52e6779ef7b038dcbb52"},
+]
+
+[package.dependencies]
+anyio = ">=4.0.0"
+mcp = ">=0.1.0"
+typing-extensions = {version = ">=4.0.0", markers = "python_version < \"3.11\""}
+
+[package.extras]
+dev = ["anyio[trio] (>=4.0.0)", "mypy (>=1.0.0)", "pytest (>=7.0.0)", "pytest-asyncio (>=0.20.0)", "pytest-cov (>=4.0.0)", "ruff (>=0.1.0)"]
+
 [[package]]
 name = "cleo"
 version = "2.1.0"
@@ -2593,6 +2616,18 @@ http2 = ["h2 (>=3,<5)"]
 socks = ["socksio (==1.*)"]
 zstd = ["zstandard (>=0.18.0)"]
 
+[[package]]
+name = "httpx-sse"
+version = "0.4.3"
+description = "Consume Server-Sent Event (SSE) messages with HTTPX."
+optional = false
+python-versions = ">=3.9"
+groups = ["main"]
+files = [
+    {file = "httpx_sse-0.4.3-py3-none-any.whl", hash = "sha256:0ac1c9fe3c0afad2e0ebb25a934a59f4c7823b60792691f779fad2c5568830fc"},
+    {file = "httpx_sse-0.4.3.tar.gz", hash = "sha256:9b1ed0127459a66014aec3c56bebd93da3c1bc8bb6618c8082039a44889a755d"},
+]
+
 [[package]]
 name = "huggingface-hub"
 version = "1.4.1"
@@ -3310,6 +3345,39 @@ files = [
     {file = "mccabe-0.7.0.tar.gz", hash = "sha256:348e0240c33b60bbdf4e523192ef919f28cb2c3d7d5c7794f74009290f236325"},
 ]
 
+[[package]]
+name = "mcp"
+version = "1.26.0"
+description = "Model Context Protocol SDK"
+optional = false
+python-versions = ">=3.10"
+groups = ["main"]
+files = [
+    {file = "mcp-1.26.0-py3-none-any.whl", hash = "sha256:904a21c33c25aa98ddbeb47273033c435e595bbacfdb177f4bd87f6dceebe1ca"},
+    {file = "mcp-1.26.0.tar.gz", hash = "sha256:db6e2ef491eecc1a0d93711a76f28dec2e05999f93afd48795da1c1137142c66"},
+]
+
+[package.dependencies]
+anyio = ">=4.5"
+httpx = ">=0.27.1"
+httpx-sse = ">=0.4"
+jsonschema = ">=4.20.0"
+pydantic = ">=2.11.0,<3.0.0"
+pydantic-settings = ">=2.5.2"
+pyjwt = {version = ">=2.10.1", extras = ["crypto"]}
+python-multipart = ">=0.0.9"
+pywin32 = {version = ">=310", markers = "sys_platform == \"win32\""}
+sse-starlette = ">=1.6.1"
+starlette = ">=0.27"
+typing-extensions = ">=4.9.0"
+typing-inspection = ">=0.4.1"
+uvicorn = {version = ">=0.31.1", markers = "sys_platform != \"emscripten\""}
+
+[package.extras]
+cli = ["python-dotenv (>=1.0.0)", "typer (>=0.16.0)"]
+rich = ["rich (>=13.9.4)"]
+ws = ["websockets (>=15.0.1)"]
+
 [[package]]
 name = "mdurl"
 version = "0.1.2"
@@ -5994,7 +6062,7 @@ description = "Python for Window Extensions"
 optional = false
 python-versions = "*"
 groups = ["main"]
-markers = "platform_system == \"Windows\""
+markers = "sys_platform == \"win32\" or platform_system == \"Windows\""
 files = [
     {file = "pywin32-311-cp310-cp310-win32.whl", hash = "sha256:d03ff496d2a0cd4a5893504789d4a15399133fe82517455e78bad62efbb7f0a3"},
     {file = "pywin32-311-cp310-cp310-win_amd64.whl", hash = "sha256:797c2772017851984b97180b0bebe4b620bb86328e8a884bb626156295a63b3b"},
@@ -6974,6 +7042,28 @@ postgresql-psycopgbinary = ["psycopg[binary] (>=3.0.7)"]
 pymysql = ["pymysql"]
 sqlcipher = ["sqlcipher3_binary"]
 
+[[package]]
+name = "sse-starlette"
+version = "3.2.0"
+description = "SSE plugin for Starlette"
+optional = false
+python-versions = ">=3.9"
+groups = ["main"]
+files = [
+    {file = "sse_starlette-3.2.0-py3-none-any.whl", hash = "sha256:5876954bd51920fc2cd51baee47a080eb88a37b5b784e615abb0b283f801cdbf"},
+    {file = "sse_starlette-3.2.0.tar.gz", hash = "sha256:8127594edfb51abe44eac9c49e59b0b01f1039d0c7461c6fd91d4e03b70da422"},
+]
+
+[package.dependencies]
+anyio = ">=4.7.0"
+starlette = ">=0.49.1"
+
+[package.extras]
+daphne = ["daphne (>=4.2.0)"]
+examples = ["aiosqlite (>=0.21.0)", "fastapi (>=0.115.12)", "sqlalchemy[asyncio] (>=2.0.41)", "uvicorn (>=0.34.0)"]
+granian = ["granian (>=2.3.1)"]
+uvicorn = ["uvicorn (>=0.34.0)"]
+
 [[package]]
 name = "stagehand"
 version = "0.5.9"
@@ -8440,4 +8530,4 @@ cffi = ["cffi (>=1.17,<2.0) ; platform_python_implementation != \"PyPy\" and pyt
 [metadata]
 lock-version = "2.1"
 python-versions = ">=3.10,<3.14"
-content-hash = "fa9c5deadf593e815dd2190f58e22152373900603f5f244b9616cd721de84d2f"
+content-hash = "55e095de555482f0fe47de7695f390fe93e7bcf739b31c391b2e5e3c3d938ae3"

autogpt_platform/backend/pyproject.toml

@@ -16,6 +16,7 @@ anthropic = "^0.79.0"
 apscheduler = "^3.11.1"
 autogpt-libs = { path = "../autogpt_libs", develop = true }
 bleach = { extras = ["css"], version = "^6.2.0" }
+claude-agent-sdk = "^0.1.0"
 click = "^8.2.0"
 cryptography = "^46.0"
 discord-py = "^2.5.2"

autogpt_platform/backend/schema.prisma

@@ -1143,154 +1143,6 @@ enum APIKeyStatus {
 
 ////////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////
-/////////////   LLM REGISTRY AND BILLING DATA   /////////////
-////////////////////////////////////////////////////////////
-////////////////////////////////////////////////////////////
-
-// LlmCostUnit: Defines how LLM MODEL costs are calculated (per run or per token).
-// This is distinct from BlockCostType (in backend/data/block.py) which defines
-// how BLOCK EXECUTION costs are calculated (per run, per byte, or per second).
-// LlmCostUnit is for pricing individual LLM model API calls in the registry,
-// while BlockCostType is for billing platform block executions.
-enum LlmCostUnit {
-  RUN
-  TOKENS
-}
-
-model LlmModelCreator {
-  id        String   @id @default(uuid())
-  createdAt DateTime @default(now())
-  updatedAt DateTime @updatedAt
-
-  name        String  @unique // e.g., "openai", "anthropic", "meta"
-  displayName String // e.g., "OpenAI", "Anthropic", "Meta"
-  description String?
-  websiteUrl  String? // Link to creator's website
-  logoUrl     String? // URL to creator's logo
-
-  metadata Json @default("{}")
-
-  Models LlmModel[]
-}
-
-model LlmProvider {
-  id        String   @id @default(uuid())
-  createdAt DateTime @default(now())
-  updatedAt DateTime @updatedAt
-
-  name        String @unique
-  displayName String
-  description String?
-
-  defaultCredentialProvider String?
-  defaultCredentialId       String?
-  defaultCredentialType     String?
-
-  supportsTools        Boolean @default(true)
-  supportsJsonOutput   Boolean @default(true)
-  supportsReasoning    Boolean @default(false)
-  supportsParallelTool Boolean @default(false)
-
-  metadata Json @default("{}")
-
-  Models LlmModel[]
-}
-
-model LlmModel {
-  id        String   @id @default(uuid())
-  createdAt DateTime @default(now())
-  updatedAt DateTime @updatedAt
-
-  slug        String @unique
-  displayName String
-  description String?
-
-  providerId String
-  Provider   LlmProvider @relation(fields: [providerId], references: [id], onDelete: Restrict)
-
-  // Creator is the organization that created/trained the model (e.g., OpenAI, Meta)
-  // This is distinct from the provider who hosts/serves the model (e.g., OpenRouter)
-  creatorId String?
-  Creator   LlmModelCreator? @relation(fields: [creatorId], references: [id], onDelete: SetNull)
-
-  contextWindow   Int
-  maxOutputTokens Int?
-  priceTier       Int     @default(1) // 1=cheapest, 2=medium, 3=expensive
-  isEnabled       Boolean @default(true)
-  isRecommended   Boolean @default(false)
-
-  capabilities Json @default("{}")
-  metadata     Json @default("{}")
-
-  Costs LlmModelCost[]
-
-  @@index([providerId, isEnabled])
-  @@index([creatorId])
-  @@index([slug])
-}
-
-model LlmModelCost {
-  id        String     @id @default(uuid())
-  createdAt DateTime   @default(now())
-  updatedAt DateTime   @updatedAt
-  unit      LlmCostUnit @default(RUN)
-
-  creditCost Int
-
-  credentialProvider String
-  credentialId       String?
-  credentialType     String?
-  currency           String?
-
-  metadata Json @default("{}")
-
-  llmModelId String
-  Model      LlmModel @relation(fields: [llmModelId], references: [id], onDelete: Cascade)
-
-  @@unique([llmModelId, credentialProvider, unit])
-  @@index([llmModelId])
-  @@index([credentialProvider])
-}
-
-// Tracks model migrations for revert capability
-// When a model is disabled with migration, we record which nodes were affected
-// so they can be reverted when the original model is back online
-model LlmModelMigration {
-  id        String   @id @default(uuid())
-  createdAt DateTime @default(now())
-  updatedAt DateTime @updatedAt
-
-  sourceModelSlug String // The original model that was disabled
-  targetModelSlug String // The model workflows were migrated to
-  reason          String? // Why the migration happened (e.g., "Provider outage")
-
-  // Track affected nodes as JSON array of node IDs
-  // Format: ["node-uuid-1", "node-uuid-2", ...]
-  migratedNodeIds Json @default("[]")
-  nodeCount       Int // Number of nodes migrated
-
-  // Custom pricing override for migrated workflows during the migration period.
-  // Use case: When migrating users from an expensive model (e.g., GPT-4) to a cheaper
-  // one (e.g., GPT-3.5), you may want to temporarily maintain the original pricing
-  // to avoid billing surprises, or offer a discount during the transition.
-  //
-  // IMPORTANT: This field is intended for integration with the billing system.
-  // When billing calculates costs for nodes affected by this migration, it should
-  // check if customCreditCost is set and use it instead of the target model's cost.
-  // If null, the target model's normal cost applies.
-  //
-  // TODO: Integrate with billing system to apply this override during cost calculation.
-  customCreditCost Int?
-
-  // Revert tracking
-  isReverted Boolean   @default(false)
-  revertedAt DateTime?
-
-  @@index([sourceModelSlug])
-  @@index([targetModelSlug])
-  @@index([isReverted])
-  @@index([sourceModelSlug, isReverted]) // Composite index for active migration queries
-}
 //////////////   OAUTH PROVIDER TABLES    //////////////////
 ////////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////

autogpt_platform/backend/test/blocks/test_jina_extract_website.py

@@ -0,0 +1,66 @@
+from typing import cast
+
+import pytest
+
+from backend.blocks.jina._auth import (
+    TEST_CREDENTIALS,
+    TEST_CREDENTIALS_INPUT,
+    JinaCredentialsInput,
+)
+from backend.blocks.jina.search import ExtractWebsiteContentBlock
+from backend.util.request import HTTPClientError
+
+
+@pytest.mark.asyncio
+async def test_extract_website_content_returns_content(monkeypatch):
+    block = ExtractWebsiteContentBlock()
+    input_data = block.Input(
+        url="https://example.com",
+        credentials=cast(JinaCredentialsInput, TEST_CREDENTIALS_INPUT),
+        raw_content=True,
+    )
+
+    async def fake_get_request(url, json=False, headers=None):
+        assert url == "https://example.com"
+        assert headers == {}
+        return "page content"
+
+    monkeypatch.setattr(block, "get_request", fake_get_request)
+
+    results = [
+        output
+        async for output in block.run(
+            input_data=input_data, credentials=TEST_CREDENTIALS
+        )
+    ]
+
+    assert ("content", "page content") in results
+    assert all(key != "error" for key, _ in results)
+
+
+@pytest.mark.asyncio
+async def test_extract_website_content_handles_http_error(monkeypatch):
+    block = ExtractWebsiteContentBlock()
+    input_data = block.Input(
+        url="https://example.com",
+        credentials=cast(JinaCredentialsInput, TEST_CREDENTIALS_INPUT),
+        raw_content=False,
+    )
+
+    async def fake_get_request(url, json=False, headers=None):
+        raise HTTPClientError("HTTP 400 Error: Bad Request", 400)
+
+    monkeypatch.setattr(block, "get_request", fake_get_request)
+
+    results = [
+        output
+        async for output in block.run(
+            input_data=input_data, credentials=TEST_CREDENTIALS
+        )
+    ]
+
+    assert ("content", "page content") not in results
+    error_messages = [value for key, value in results if key == "error"]
+    assert error_messages
+    assert "Client error (400)" in error_messages[0]
+    assert "https://example.com" in error_messages[0]