Clean up mcp

apps/sim/app/api/mcp/copilot/route.ts

@@ -34,144 +34,40 @@ export const dynamic = 'force-dynamic'
  * the workflow lifecycle and best practices.
  */
 const MCP_SERVER_INSTRUCTIONS = `
-## Sim Workflow Copilot - Usage Guide
+## Sim Workflow Copilot
 
-You are interacting with Sim's workflow automation platform. These tools orchestrate specialized AI agents that build workflows. Follow these guidelines carefully.
+Sim is a workflow automation platform. Workflows are visual pipelines of connected blocks (Agent, Function, Condition, API, integrations, etc.). The Agent block is the core — an LLM with tools, memory, structured output, and knowledge bases.
 
----
+### Workflow Lifecycle (Happy Path)
 
-## Platform Knowledge
+1. \`list_workspaces\` → know where to work
+2. \`create_workflow(name, workspaceId)\` → get a workflowId
+3. \`copilot_build(request, workflowId)\` → plan and build in one pass
+4. \`copilot_test(request, workflowId)\` → verify it works
+5. \`copilot_deploy("deploy as api", workflowId)\` → make it accessible externally (optional)
 
-Sim is a workflow automation platform. Workflows are visual pipelines of blocks.
+For fine-grained control, use \`copilot_plan\` → \`copilot_edit\` instead of \`copilot_build\`. Pass the plan object from copilot_plan EXACTLY as-is to copilot_edit's context.plan field.
 
-### Block Types
+### Working with Existing Workflows
 
-**Core Logic:**
-- **Agent** - The heart of Sim (LLM block with tools, memory, structured output, knowledge bases)
-- **Function** - JavaScript code execution
-- **Condition** - If/else branching
-- **Router** - AI-powered content-based routing
-- **Loop** - While/do-while iteration
-- **Parallel** - Simultaneous execution
-- **API** - HTTP requests
+When the user refers to a workflow by name or description ("the email one", "my Slack bot"):
+1. Use \`copilot_discovery\` to find it by functionality
+2. Or use \`list_workflows\` and match by name
+3. Then pass the workflowId to other tools
 
-**Integrations (3rd Party):**
-- OAuth: Slack, Gmail, Google Calendar, Sheets, Outlook, Linear, GitHub, Notion
-- API: Stripe, Twilio, SendGrid, any REST API
+### Organization
 
-### The Agent Block
+- \`rename_workflow\` — rename a workflow
+- \`move_workflow\` — move a workflow into a folder (or root with null)
+- \`move_folder\` — nest a folder inside another (or root with null)
+- \`create_folder(name, parentId)\` — create nested folder hierarchies
 
-The Agent block is the core of intelligent workflows:
-- **Tools** - Add integrations, custom tools, web search to give it capabilities
-- **Memory** - Multi-turn conversations with persistent context
-- **Structured Output** - JSON schema for reliable parsing
-- **Knowledge Bases** - RAG-powered document retrieval
+### Key Rules
 
-**Design principle:** Put tools INSIDE agents rather than using standalone tool blocks.
-
-### Triggers
-
-| Type | Description |
-|------|-------------|
-| Manual/Chat | User sends message in UI (start block: input, files, conversationId) |
-| API | REST endpoint with custom input schema |
-| Webhook | External services POST to trigger URL |
-| Schedule | Cron-based (hourly, daily, weekly) |
-
-### Deployments
-
-| Type | Trigger | Use Case |
-|------|---------|----------|
-| API | Start block | REST endpoint for programmatic access |
-| Chat | Start block | Managed chat UI with auth options |
-| MCP | Start block | Expose as MCP tool for AI agents |
-| General | Schedule/Webhook | Activate triggers to run automatically |
-
-**Undeployed workflows only run in the builder UI.**
-
-### Variable Syntax
-
-Reference outputs from previous blocks: \`<blockname.field>\`
-Reference environment variables: \`{{ENV_VAR_NAME}}\`
-
-Rules:
-- Block names must be lowercase, no spaces, no special characters
-- Use dot notation for nested fields: \`<blockname.field.subfield>\`
-
----
-
-## Workflow Lifecycle
-
-1. **Create**: For NEW workflows, FIRST call create_workflow to get a workflowId
-2. **Plan**: Use copilot_plan with the workflowId to plan the workflow
-3. **Edit**: Use copilot_edit with the workflowId AND the plan to build the workflow
-4. **Deploy**: ALWAYS deploy after building using copilot_deploy before testing/running
-5. **Test**: Use copilot_test to verify the workflow works correctly
-6. **Share**: Provide the user with the workflow URL after completion
-
----
-
-## CRITICAL: Always Pass workflowId
-
-- For NEW workflows: Call create_workflow FIRST, then use the returned workflowId
-- For EXISTING workflows: Pass the workflowId to all copilot tools
-- copilot_plan, copilot_edit, copilot_deploy, copilot_test, copilot_debug all REQUIRE workflowId
-
----
-
-## CRITICAL: How to Handle Plans
-
-The copilot_plan tool returns a structured plan object. You MUST:
-
-1. **Do NOT modify the plan**: Pass the plan object EXACTLY as returned to copilot_edit
-2. **Do NOT interpret or summarize the plan**: The edit agent needs the raw plan data
-3. **Pass the plan in the context.plan field**: \`{ "context": { "plan": <plan_object> } }\`
-4. **Include ALL plan data**: Block configurations, connections, credentials, everything
-
-Example flow:
-\`\`\`
-1. copilot_plan({ request: "build a workflow...", workflowId: "abc123" })
-   -> Returns: { "plan": { "blocks": [...], "connections": [...], ... } }
-
-2. copilot_edit({ 
-     workflowId: "abc123",
-     message: "Execute the plan",
-     context: { "plan": <EXACT plan object from step 1> }
-   })
-\`\`\`
-
-**Why this matters**: The plan contains technical details (block IDs, field mappings, API schemas) that the edit agent needs verbatim. Summarizing or rephrasing loses critical information.
-
----
-
-## CRITICAL: Error Handling
-
-**If the user says "doesn't work", "broke", "failed", "error" → ALWAYS use copilot_debug FIRST.**
-
-Don't guess. Don't plan. Debug first to find the actual problem.
-
----
-
-## Important Rules
-
-- ALWAYS deploy a workflow before attempting to run or test it
-- Workflows must be deployed to have an "active deployment" for execution
-- After building, call copilot_deploy with the appropriate deployment type (api, chat, or mcp)
-- Return the workflow URL to the user so they can access it in Sim
-
----
-
-## Quick Operations (use direct tools)
-- list_workflows, list_workspaces, list_folders, get_workflow: Fast database queries
-- create_workflow: Create new workflow and get workflowId (CALL THIS FIRST for new workflows)
-- create_folder: Create new resources
-
-## Workflow Building (use copilot tools)
-- copilot_plan: Plan workflow changes (REQUIRES workflowId) - returns a plan object
-- copilot_edit: Execute the plan (REQUIRES workflowId AND plan from copilot_plan)
-- copilot_deploy: Deploy workflows (REQUIRES workflowId)
-- copilot_test: Test workflow execution (REQUIRES workflowId)
-- copilot_debug: Diagnose errors (REQUIRES workflowId) - USE THIS FIRST for issues
+- You can test workflows immediately after building — deployment is only needed for external access (API, chat, MCP).
+- All copilot tools (build, plan, edit, deploy, test, debug) require workflowId.
+- If the user reports errors → use \`copilot_debug\` first, don't guess.
+- Variable syntax: \`<blockname.field>\` for block outputs, \`{{ENV_VAR}}\` for env vars.
 `
 
 function createResponse(id: RequestId, result: unknown): JSONRPCResponse {
@@ -378,7 +274,6 @@ async function handleBuildToolCall(
     }
 
     const chatId = crypto.randomUUID()
-    const context = (args.context as Record<string, unknown>) || {}
 
     const requestPayload = {
       message: requestText,
@@ -391,7 +286,6 @@ async function handleBuildToolCall(
       version: SIM_AGENT_VERSION,
       headless: true,
       chatId,
-      context,
     }
 
     const result = await orchestrateCopilotStream(requestPayload, {

apps/sim/lib/copilot/orchestrator/tool-executor/index.ts

@@ -38,6 +38,9 @@ import type {
   ListFoldersParams,
   ListUserWorkflowsParams,
   ListWorkspaceMcpServersParams,
+  MoveFolderParams,
+  MoveWorkflowParams,
+  RenameWorkflowParams,
   RunWorkflowParams,
   SetGlobalWorkflowVariablesParams,
 } from './param-types'
@@ -52,6 +55,9 @@ import {
   executeListFolders,
   executeListUserWorkflows,
   executeListUserWorkspaces,
+  executeMoveFolder,
+  executeMoveWorkflow,
+  executeRenameWorkflow,
   executeRunWorkflow,
   executeSetGlobalWorkflowVariables,
 } from './workflow-tools'
@@ -85,6 +91,9 @@ const SIM_WORKFLOW_TOOL_HANDLERS: Record<
   list_folders: (p, c) => executeListFolders(p as ListFoldersParams, c),
   create_workflow: (p, c) => executeCreateWorkflow(p as CreateWorkflowParams, c),
   create_folder: (p, c) => executeCreateFolder(p as CreateFolderParams, c),
+  rename_workflow: (p, c) => executeRenameWorkflow(p as unknown as RenameWorkflowParams, c),
+  move_workflow: (p, c) => executeMoveWorkflow(p as unknown as MoveWorkflowParams, c),
+  move_folder: (p, c) => executeMoveFolder(p as unknown as MoveFolderParams, c),
   get_workflow_data: (p, c) => executeGetWorkflowData(p as GetWorkflowDataParams, c),
   get_block_outputs: (p, c) => executeGetBlockOutputs(p as GetBlockOutputsParams, c),
   get_block_upstream_references: (p, c) =>

apps/sim/lib/copilot/orchestrator/tool-executor/param-types.ts

@@ -125,3 +125,20 @@ export interface CreateWorkspaceMcpServerParams {
   isPublic?: boolean
   workflowIds?: string[]
 }
+
+// === Workflow Organization Params ===
+
+export interface RenameWorkflowParams {
+  workflowId: string
+  name: string
+}
+
+export interface MoveWorkflowParams {
+  workflowId: string
+  folderId: string | null
+}
+
+export interface MoveFolderParams {
+  folderId: string
+  parentId: string | null
+}

apps/sim/lib/copilot/orchestrator/tool-executor/workflow-tools/mutations.ts

@@ -12,6 +12,9 @@ import { ensureWorkflowAccess, ensureWorkspaceAccess, getDefaultWorkspaceId } fr
 import type {
   CreateFolderParams,
   CreateWorkflowParams,
+  MoveFolderParams,
+  MoveWorkflowParams,
+  RenameWorkflowParams,
   RunWorkflowParams,
   SetGlobalWorkflowVariablesParams,
   VariableOperation,
@@ -283,3 +286,85 @@ export async function executeSetGlobalWorkflowVariables(
     return { success: false, error: error instanceof Error ? error.message : String(error) }
   }
 }
+
+export async function executeRenameWorkflow(
+  params: RenameWorkflowParams,
+  context: ExecutionContext
+): Promise<ToolCallResult> {
+  try {
+    const workflowId = params.workflowId
+    if (!workflowId) {
+      return { success: false, error: 'workflowId is required' }
+    }
+    const name = typeof params.name === 'string' ? params.name.trim() : ''
+    if (!name) {
+      return { success: false, error: 'name is required' }
+    }
+    if (name.length > 200) {
+      return { success: false, error: 'Workflow name must be 200 characters or less' }
+    }
+
+    await ensureWorkflowAccess(workflowId, context.userId)
+
+    await db
+      .update(workflow)
+      .set({ name, updatedAt: new Date() })
+      .where(eq(workflow.id, workflowId))
+
+    return { success: true, output: { workflowId, name } }
+  } catch (error) {
+    return { success: false, error: error instanceof Error ? error.message : String(error) }
+  }
+}
+
+export async function executeMoveWorkflow(
+  params: MoveWorkflowParams,
+  context: ExecutionContext
+): Promise<ToolCallResult> {
+  try {
+    const workflowId = params.workflowId
+    if (!workflowId) {
+      return { success: false, error: 'workflowId is required' }
+    }
+
+    await ensureWorkflowAccess(workflowId, context.userId)
+
+    const folderId = params.folderId || null
+
+    await db
+      .update(workflow)
+      .set({ folderId, updatedAt: new Date() })
+      .where(eq(workflow.id, workflowId))
+
+    return { success: true, output: { workflowId, folderId } }
+  } catch (error) {
+    return { success: false, error: error instanceof Error ? error.message : String(error) }
+  }
+}
+
+export async function executeMoveFolder(
+  params: MoveFolderParams,
+  context: ExecutionContext
+): Promise<ToolCallResult> {
+  try {
+    const folderId = params.folderId
+    if (!folderId) {
+      return { success: false, error: 'folderId is required' }
+    }
+
+    const parentId = params.parentId || null
+
+    if (parentId === folderId) {
+      return { success: false, error: 'A folder cannot be moved into itself' }
+    }
+
+    await db
+      .update(workflowFolder)
+      .set({ parentId, updatedAt: new Date() })
+      .where(eq(workflowFolder.id, folderId))
+
+    return { success: true, output: { folderId, parentId } }
+  } catch (error) {
+    return { success: false, error: error instanceof Error ? error.message : String(error) }
+  }
+}

apps/sim/lib/copilot/orchestrator/tool-executor/workflow-tools/queries.ts

@@ -3,7 +3,6 @@ import { customTools, permissions, workflow, workflowFolder, workspace } from '@
 import { and, asc, desc, eq, isNull, or } from 'drizzle-orm'
 import type { ExecutionContext, ToolCallResult } from '@/lib/copilot/orchestrator/types'
 import {
-  extractWorkflowNames,
   formatNormalizedWorkflowForCopilot,
   normalizeWorkflowName,
 } from '@/lib/copilot/tools/shared/workflow-utils'
@@ -114,8 +113,6 @@ export async function executeListUserWorkflows(
 
     const workflows = await getAccessibleWorkflowsForUser(context.userId, { workspaceId, folderId })
 
-    const names = extractWorkflowNames(workflows)
-
     const workflowList = workflows.map((w) => ({
       workflowId: w.id,
       workflowName: w.name || '',
@@ -123,7 +120,7 @@ export async function executeListUserWorkflows(
       folderId: w.folderId,
     }))
 
-    return { success: true, output: { workflow_names: names, workflows: workflowList } }
+    return { success: true, output: { workflows: workflowList } }
   } catch (error) {
     return { success: false, error: error instanceof Error ? error.message : String(error) }
   }

apps/sim/lib/copilot/tools/mcp/definitions.ts

@@ -17,11 +17,21 @@ export type SubagentToolDef = {
  * These are fast database queries that don't need AI reasoning.
  */
 export const DIRECT_TOOL_DEFS: DirectToolDef[] = [
+  {
+    name: 'list_workspaces',
+    toolId: 'list_user_workspaces',
+    description:
+      'List all workspaces the user has access to. Returns workspace IDs, names, and roles. Use this first to determine which workspace to operate in.',
+    inputSchema: {
+      type: 'object',
+      properties: {},
+    },
+  },
   {
     name: 'list_workflows',
     toolId: 'list_user_workflows',
     description:
-      'List all workflows the user has access to. Returns workflow IDs, names, and workspace info.',
+      'List all workflows the user has access to. Returns workflow IDs, names, workspace, and folder info. Use workspaceId/folderId to scope results.',
     inputSchema: {
       type: 'object',
       properties: {
@@ -36,20 +46,11 @@ export const DIRECT_TOOL_DEFS: DirectToolDef[] = [
       },
     },
   },
-  {
-    name: 'list_workspaces',
-    toolId: 'list_user_workspaces',
-    description:
-      'List all workspaces the user has access to. Returns workspace IDs, names, and roles.',
-    inputSchema: {
-      type: 'object',
-      properties: {},
-    },
-  },
   {
     name: 'list_folders',
     toolId: 'list_folders',
-    description: 'List all folders in a workspace.',
+    description:
+      'List all folders in a workspace. Returns folder IDs, names, and parent relationships for organizing workflows.',
     inputSchema: {
       type: 'object',
       properties: {
@@ -64,7 +65,8 @@ export const DIRECT_TOOL_DEFS: DirectToolDef[] = [
   {
     name: 'get_workflow',
     toolId: 'get_user_workflow',
-    description: 'Get a workflow by ID. Returns the full workflow definition.',
+    description:
+      'Get a workflow by ID. Returns the full workflow definition including all blocks, connections, and configuration.',
     inputSchema: {
       type: 'object',
       properties: {
@@ -79,7 +81,8 @@ export const DIRECT_TOOL_DEFS: DirectToolDef[] = [
   {
     name: 'create_workflow',
     toolId: 'create_workflow',
-    description: 'Create a new workflow. Returns the new workflow ID.',
+    description:
+      'Create a new empty workflow. Returns the new workflow ID. Always call this FIRST before copilot_build for new workflows. Use workspaceId to place it in a specific workspace.',
     inputSchema: {
       type: 'object',
       properties: {
@@ -106,7 +109,8 @@ export const DIRECT_TOOL_DEFS: DirectToolDef[] = [
   {
     name: 'create_folder',
     toolId: 'create_folder',
-    description: 'Create a new folder in a workspace.',
+    description:
+      'Create a new folder for organizing workflows. Use parentId to create nested folder hierarchies.',
     inputSchema: {
       type: 'object',
       properties: {
@@ -126,6 +130,65 @@ export const DIRECT_TOOL_DEFS: DirectToolDef[] = [
       required: ['name'],
     },
   },
+  {
+    name: 'rename_workflow',
+    toolId: 'rename_workflow',
+    description: 'Rename an existing workflow.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        workflowId: {
+          type: 'string',
+          description: 'The workflow ID to rename.',
+        },
+        name: {
+          type: 'string',
+          description: 'The new name for the workflow.',
+        },
+      },
+      required: ['workflowId', 'name'],
+    },
+  },
+  {
+    name: 'move_workflow',
+    toolId: 'move_workflow',
+    description:
+      'Move a workflow into a different folder. Set folderId to null to move to the workspace root.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        workflowId: {
+          type: 'string',
+          description: 'The workflow ID to move.',
+        },
+        folderId: {
+          type: ['string', 'null'],
+          description: 'Target folder ID, or null to move to workspace root.',
+        },
+      },
+      required: ['workflowId', 'folderId'],
+    },
+  },
+  {
+    name: 'move_folder',
+    toolId: 'move_folder',
+    description:
+      'Move a folder into another folder. Set parentId to null to move to the workspace root.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        folderId: {
+          type: 'string',
+          description: 'The folder ID to move.',
+        },
+        parentId: {
+          type: ['string', 'null'],
+          description: 'Target parent folder ID, or null to move to workspace root.',
+        },
+      },
+      required: ['folderId', 'parentId'],
+    },
+  },
 ]
 
 export const SUBAGENT_TOOL_DEFS: SubagentToolDef[] = [
@@ -151,15 +214,15 @@ CAN DO:
 - Set environment variables and workflow variables
 
 CANNOT DO:
-- Run or test workflows (use copilot_test separately after deploying)
+- Run or test workflows (use copilot_test separately)
 - Deploy workflows (use copilot_deploy separately)
 
 WORKFLOW:
 1. Call create_workflow to get a workflowId (for new workflows)
 2. Call copilot_build with the request and workflowId
 3. Build agent gathers info and builds in one pass
-4. Call copilot_deploy to deploy the workflow
-5. Optionally call copilot_test to verify it works`,
+4. Call copilot_test to verify it works
+5. Optionally call copilot_deploy to make it externally accessible`,
     inputSchema: {
       type: 'object',
       properties: {
@@ -205,13 +268,11 @@ DO NOT USE (use direct tools instead):
   {
     name: 'copilot_plan',
     agentId: 'plan',
-    description: `Plan workflow changes by gathering required information.
+    description: `Plan workflow changes by gathering required information. For most cases, prefer copilot_build which combines planning and editing in one step.
 
 USE THIS WHEN:
-- Building a new workflow
-- Modifying an existing workflow
-- You need to understand what blocks and integrations are available
-- The workflow requires multiple blocks or connections
+- You need fine-grained control over the build process
+- You want to inspect the plan before executing it
 
 WORKFLOW ID (REQUIRED):
 - For NEW workflows: First call create_workflow to get a workflowId, then pass it here
@@ -241,23 +302,16 @@ IMPORTANT: Pass the returned plan EXACTLY to copilot_edit - do not modify or sum
   {
     name: 'copilot_edit',
     agentId: 'edit',
-    description: `Execute a workflow plan and apply edits.
-
-USE THIS WHEN:
-- You have a plan from copilot_plan that needs to be executed
-- Building or modifying a workflow based on the plan
-- Making changes to blocks, connections, or configurations
+    description: `Execute a workflow plan from copilot_plan. For most cases, prefer copilot_build which combines planning and editing in one step.
 
 WORKFLOW ID (REQUIRED):
 - You MUST provide the workflowId parameter
-- For new workflows, get the workflowId from create_workflow first
 
 PLAN (REQUIRED):
 - Pass the EXACT plan object from copilot_plan in the context.plan field
 - Do NOT modify, summarize, or interpret the plan - pass it verbatim
-- The plan contains technical details the edit agent needs exactly as-is
 
-IMPORTANT: After copilot_edit completes, you MUST call copilot_deploy before the workflow can be run or tested.`,
+After copilot_edit completes, you can test immediately with copilot_test, or deploy with copilot_deploy to make it accessible externally.`,
     inputSchema: {
       type: 'object',
       properties: {
@@ -281,43 +335,15 @@ IMPORTANT: After copilot_edit completes, you MUST call copilot_deploy before the
       required: ['workflowId'],
     },
   },
-  {
-    name: 'copilot_debug',
-    agentId: 'debug',
-    description: `Diagnose errors or unexpected workflow behavior.
-
-WORKFLOW ID (REQUIRED): Always provide the workflowId of the workflow to debug.`,
-    inputSchema: {
-      type: 'object',
-      properties: {
-        error: { type: 'string', description: 'The error message or description of the issue.' },
-        workflowId: { type: 'string', description: 'REQUIRED. The workflow ID to debug.' },
-        context: { type: 'object' },
-      },
-      required: ['error', 'workflowId'],
-    },
-  },
   {
     name: 'copilot_deploy',
     agentId: 'deploy',
-    description: `Deploy or manage workflow deployments.
-
-CRITICAL: You MUST deploy a workflow after building before it can be run or tested.
-Workflows without an active deployment will fail with "no active deployment" error.
-
-WORKFLOW ID (REQUIRED):
-- Always provide the workflowId parameter
-- This must match the workflow you built with copilot_edit
-
-USE THIS:
-- After copilot_edit completes to activate the workflow
-- To update deployment settings
-- To redeploy after making changes
+    description: `Deploy a workflow to make it accessible externally. Workflows can be tested without deploying, but deployment is needed for API access, chat UIs, or MCP exposure.
 
 DEPLOYMENT TYPES:
-- "deploy as api" - REST API endpoint
-- "deploy as chat" - Chat interface
-- "deploy as mcp" - MCP server`,
+- "deploy as api" - REST API endpoint for programmatic access
+- "deploy as chat" - Managed chat UI with auth options
+- "deploy as mcp" - Expose as MCP tool for AI agents`,
     inputSchema: {
       type: 'object',
       properties: {
@@ -334,114 +360,10 @@ DEPLOYMENT TYPES:
       required: ['request', 'workflowId'],
     },
   },
-  {
-    name: 'copilot_auth',
-    agentId: 'auth',
-    description: 'Handle OAuth connection flows.',
-    inputSchema: {
-      type: 'object',
-      properties: {
-        request: { type: 'string' },
-        context: { type: 'object' },
-      },
-      required: ['request'],
-    },
-  },
-  {
-    name: 'copilot_knowledge',
-    agentId: 'knowledge',
-    description: 'Create and manage knowledge bases.',
-    inputSchema: {
-      type: 'object',
-      properties: {
-        request: { type: 'string' },
-        context: { type: 'object' },
-      },
-      required: ['request'],
-    },
-  },
-  {
-    name: 'copilot_custom_tool',
-    agentId: 'custom_tool',
-    description: 'Create or manage custom tools.',
-    inputSchema: {
-      type: 'object',
-      properties: {
-        request: { type: 'string' },
-        context: { type: 'object' },
-      },
-      required: ['request'],
-    },
-  },
-  {
-    name: 'copilot_info',
-    agentId: 'info',
-    description: 'Inspect blocks, outputs, and workflow metadata.',
-    inputSchema: {
-      type: 'object',
-      properties: {
-        request: { type: 'string' },
-        workflowId: { type: 'string' },
-        context: { type: 'object' },
-      },
-      required: ['request'],
-    },
-  },
-  {
-    name: 'copilot_workflow',
-    agentId: 'workflow',
-    description: 'Manage workflow environment and configuration.',
-    inputSchema: {
-      type: 'object',
-      properties: {
-        request: { type: 'string' },
-        workflowId: { type: 'string' },
-        context: { type: 'object' },
-      },
-      required: ['request'],
-    },
-  },
-  {
-    name: 'copilot_research',
-    agentId: 'research',
-    description: 'Research external APIs and documentation.',
-    inputSchema: {
-      type: 'object',
-      properties: {
-        request: { type: 'string' },
-        context: { type: 'object' },
-      },
-      required: ['request'],
-    },
-  },
-  {
-    name: 'copilot_tour',
-    agentId: 'tour',
-    description: 'Explain platform features and usage.',
-    inputSchema: {
-      type: 'object',
-      properties: {
-        request: { type: 'string' },
-        context: { type: 'object' },
-      },
-      required: ['request'],
-    },
-  },
   {
     name: 'copilot_test',
     agentId: 'test',
-    description: `Run workflows and verify outputs.
-
-PREREQUISITE: The workflow MUST be deployed first using copilot_deploy.
-Undeployed workflows will fail with "no active deployment" error.
-
-WORKFLOW ID (REQUIRED):
-- Always provide the workflowId parameter
-
-USE THIS:
-- After deploying to verify the workflow works correctly
-- To test with sample inputs
-- To validate workflow behavior before sharing with user`,
+    description: `Run a workflow and verify its outputs. Works on both deployed and undeployed (draft) workflows. Use after building to verify correctness.`,
     inputSchema: {
       type: 'object',
       properties: {
@@ -455,10 +377,126 @@ USE THIS:
       required: ['request', 'workflowId'],
     },
   },
+  {
+    name: 'copilot_debug',
+    agentId: 'debug',
+    description:
+      'Diagnose errors or unexpected workflow behavior. Provide the error message and workflowId. Returns root cause analysis and fix suggestions.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        error: { type: 'string', description: 'The error message or description of the issue.' },
+        workflowId: { type: 'string', description: 'REQUIRED. The workflow ID to debug.' },
+        context: { type: 'object' },
+      },
+      required: ['error', 'workflowId'],
+    },
+  },
+  {
+    name: 'copilot_auth',
+    agentId: 'auth',
+    description:
+      'Check OAuth connection status, list connected services, and initiate new OAuth connections. Use when a workflow needs third-party service access (Google, Slack, GitHub, etc.).',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        request: { type: 'string' },
+        context: { type: 'object' },
+      },
+      required: ['request'],
+    },
+  },
+  {
+    name: 'copilot_knowledge',
+    agentId: 'knowledge',
+    description:
+      'Manage knowledge bases for RAG-powered document retrieval. Supports listing, creating, updating, and deleting knowledge bases. Knowledge bases can be attached to agent blocks for context-aware responses.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        request: { type: 'string' },
+        context: { type: 'object' },
+      },
+      required: ['request'],
+    },
+  },
+  {
+    name: 'copilot_custom_tool',
+    agentId: 'custom_tool',
+    description:
+      'Manage custom tools (reusable API integrations). Supports listing, creating, updating, and deleting custom tools. Custom tools can be added to agent blocks as callable functions.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        request: { type: 'string' },
+        context: { type: 'object' },
+      },
+      required: ['request'],
+    },
+  },
+  {
+    name: 'copilot_info',
+    agentId: 'info',
+    description:
+      'Inspect a workflow\'s blocks, connections, outputs, variables, and metadata. Always provide workflowId to scope results to a specific workflow.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        request: { type: 'string' },
+        workflowId: { type: 'string' },
+        context: { type: 'object' },
+      },
+      required: ['request'],
+    },
+  },
+  {
+    name: 'copilot_workflow',
+    agentId: 'workflow',
+    description:
+      'Manage workflow-level configuration: environment variables, settings, scheduling, and deployment status.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        request: { type: 'string' },
+        workflowId: { type: 'string' },
+        context: { type: 'object' },
+      },
+      required: ['request'],
+    },
+  },
+  {
+    name: 'copilot_research',
+    agentId: 'research',
+    description:
+      'Research external APIs and documentation. Use when building workflows that integrate with third-party services and you need to understand their API, authentication, or data formats.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        request: { type: 'string' },
+        context: { type: 'object' },
+      },
+      required: ['request'],
+    },
+  },
+  {
+    name: 'copilot_tour',
+    agentId: 'tour',
+    description:
+      'Explain platform features, concepts, and usage patterns. Use when the user asks "how does X work?" about the Sim platform, block types, triggers, deployments, or workflow concepts.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        request: { type: 'string' },
+        context: { type: 'object' },
+      },
+      required: ['request'],
+    },
+  },
   {
     name: 'copilot_superagent',
     agentId: 'superagent',
-    description: 'Execute direct external actions (email, Slack, etc.).',
+    description:
+      'Execute direct actions NOW: send an email, post to Slack, make an API call, etc. Use when the user wants to DO something immediately rather than build a workflow for it.',
     inputSchema: {
       type: 'object',
       properties: {