mirror of
https://github.com/googleapis/genai-toolbox.git
synced 2026-01-19 12:28:06 -05:00
Compare commits
5 Commits
migrate-go
...
docs-py-sd
| Author | SHA1 | Date | |
|---|---|---|---|
|
779128903c | ||
|
|
f0c7eb129b | ||
|
|
fef07c71a1 | ||
|
|
12b25a0beb | ||
|
|
073c8b3268 |
@@ -7,41 +7,6 @@ description: >
|
||||
---
|
||||
|
||||
|
||||
## Overview
|
||||
|
||||
|
||||
|
||||
# MCP Toolbox SDKs for Go
|
||||
|
||||
[](https://opensource.org/licenses/Apache-2.0)
|
||||
[](https://googleapis.github.io/genai-toolbox/)
|
||||
[](https://discord.gg/Dmm69peqjh)
|
||||
[](https://medium.com/@mcp_toolbox)
|
||||
[](https://goreportcard.com/report/github.com/googleapis/mcp-toolbox-sdk-go)
|
||||
[](https://img.shields.io/github/v/release/googleapis/mcp-toolbox-sdk-go)
|
||||
[](https://img.shields.io/github/go-mod/go-version/googleapis/mcp-toolbox-sdk-go)
|
||||
|
||||
This repository contains the Go SDK designed to seamlessly integrate the
|
||||
functionalities of the [MCP
|
||||
Toolbox](https://github.com/googleapis/genai-toolbox) into your Gen AI
|
||||
applications. The SDK allow you to load tools defined in Toolbox and use them
|
||||
as standard Go tools within popular orchestration frameworks
|
||||
or your custom code.
|
||||
|
||||
This simplifies the process of incorporating external functionalities (like
|
||||
Databases or APIs) managed by Toolbox into your GenAI applications.
|
||||
|
||||
<!-- TOC -->
|
||||
|
||||
- [Overview](#overview)
|
||||
- [Which Package Should I Use?](#which-package-should-i-use)
|
||||
- [Available Packages](#available-packages)
|
||||
- [Getting Started](#getting-started)
|
||||
|
||||
<!-- /TOC -->
|
||||
|
||||
## Overview
|
||||
|
||||
The MCP Toolbox service provides a centralized way to manage and expose tools
|
||||
@@ -57,58 +22,4 @@ The Go SDK act as clients for that service. They handle the communication needed
|
||||
By using the SDK, you can easily leverage your Toolbox-managed tools directly
|
||||
within your Go applications or AI orchestration frameworks.
|
||||
|
||||
## Which Package Should I Use?
|
||||
|
||||
Choosing the right package depends on how you are building your application:
|
||||
|
||||
- [`core`](https://github.com/googleapis/mcp-toolbox-sdk-go/tree/main/core):
|
||||
This is a framework agnostic way to connect the tools to popular frameworks
|
||||
like Google GenAI, LangChain, etc.
|
||||
|
||||
- [`tbadk`](https://github.com/googleapis/mcp-toolbox-sdk-go/tree/main/tbadk):
|
||||
This package provides a way to connect tools to ADK Go.
|
||||
|
||||
- [`tbgenkit`](https://github.com/googleapis/mcp-toolbox-sdk-go/tree/main/tbgenkit):
|
||||
This package provides a functionality to convert the Tool fetched using the core package
|
||||
into a Genkit Go compatible tool.
|
||||
|
||||
## Available Packages
|
||||
|
||||
This repository hosts the following Go packages. See the package-specific
|
||||
README for detailed installation and usage instructions:
|
||||
|
||||
| Package | Target Use Case | Integration | Path | Details (README) |
|
||||
| :------ | :----------| :---------- | :---------------------- | :---------- |
|
||||
| `core` | Framework-agnostic / Custom applications | Use directly / Custom | `core/` | 📄 [View README](https://github.com/googleapis/mcp-toolbox-sdk-go/blob/main/core/README.md) |
|
||||
| `tbadk` | ADK Go | Use directly | `tbadk/` | 📄 [View README](https://github.com/googleapis/mcp-toolbox-sdk-go/blob/main/tbadk/README.md) |
|
||||
| `tbgenkit` | Genkit Go | Along with core | `tbgenkit/` | 📄 [View README](https://github.com/googleapis/mcp-toolbox-sdk-go/blob/main/tbgenkit/README.md) |
|
||||
|
||||
## Getting Started
|
||||
|
||||
To get started using Toolbox tools with an application, follow these general steps:
|
||||
|
||||
1. **Set up and Run the Toolbox Service:**
|
||||
|
||||
Before using the SDKs, you need the MCP Toolbox server running. Follow
|
||||
the instructions here: [**Toolbox Getting Started
|
||||
Guide**](https://github.com/googleapis/genai-toolbox?tab=readme-ov-file#getting-started)
|
||||
|
||||
2. **Install the Appropriate SDK:**
|
||||
|
||||
Choose the package based on your needs (see "[Which Package Should I Use?](#which-package-should-i-use)" above)
|
||||
Use this command to install the SDK module
|
||||
|
||||
```bash
|
||||
# For the core, framework-agnostic SDK
|
||||
go get github.com/googleapis/mcp-toolbox-sdk-go
|
||||
```
|
||||
|
||||
3. **Use the SDK:**
|
||||
|
||||
Consult the README for your chosen package (linked in the "[Available
|
||||
Packages](#available-packages)" section above) for detailed instructions on
|
||||
how to connect the client, load tool definitions, invoke tools, configure
|
||||
authentication/binding, and integrate them into your application or
|
||||
framework.
|
||||
|
||||
[Github](https://github.com/googleapis/mcp-toolbox-sdk-go)
|
||||
@@ -1,998 +0,0 @@
|
||||
---
|
||||
title: "Core Package"
|
||||
linkTitle: "Core"
|
||||
type: docs
|
||||
weight: 1
|
||||
---
|
||||
|
||||
|
||||
|
||||
# MCP Toolbox Core SDK
|
||||
|
||||
[](https://opensource.org/licenses/Apache-2.0)
|
||||
|
||||
This SDK allows you to seamlessly integrate the functionalities of
|
||||
[Toolbox](https://github.com/googleapis/genai-toolbox) allowing you to load and
|
||||
use tools defined in the service as standard Go structs within your GenAI
|
||||
applications.
|
||||
|
||||
This simplifies integrating external functionalities (like APIs, databases, or
|
||||
custom logic) managed by the Toolbox into your workflows, especially those
|
||||
involving Large Language Models (LLMs).
|
||||
|
||||
<!-- TOC ignore:true -->
|
||||
<!-- TOC -->
|
||||
|
||||
- [MCP Toolbox Core SDK](#mcp-toolbox-core-sdk)
|
||||
- [Installation](#installation)
|
||||
- [Quickstart](#quickstart)
|
||||
- [Usage](#usage)
|
||||
- [Transport Protocols](#transport-protocols)
|
||||
- [Supported Protocols](#supported-protocols)
|
||||
- [Example](#example)
|
||||
- [Loading Tools](#loading-tools)
|
||||
- [Load a toolset](#load-a-toolset)
|
||||
- [Load a single tool](#load-a-single-tool)
|
||||
- [Invoking Tools](#invoking-tools)
|
||||
- [Client to Server Authentication](#client-to-server-authentication)
|
||||
- [When is Client-to-Server Authentication Needed?](#when-is-client-to-server-authentication-needed)
|
||||
- [How it works](#how-it-works)
|
||||
- [Configuration](#configuration)
|
||||
- [Authenticating with Google Cloud Servers](#authenticating-with-google-cloud-servers)
|
||||
- [Step by Step Guide for Cloud Run](#step-by-step-guide-for-cloud-run)
|
||||
- [Authenticating Tools](#authenticating-tools)
|
||||
- [When is Authentication Needed?](#when-is-authentication-needed)
|
||||
- [Supported Authentication Mechanisms](#supported-authentication-mechanisms)
|
||||
- [Step 1: Configure Tools in Toolbox Service](#step-1-configure-tools-in-toolbox-service)
|
||||
- [Step 2: Configure SDK Client](#step-2-configure-sdk-client)
|
||||
- [Provide an ID Token Retriever Function](#provide-an-id-token-retriever-function)
|
||||
- [Option A: Add Default Authentication to a Client](#option-a-add-default-authentication-to-a-client)
|
||||
- [Option B: Add Authentication to a Loaded Tool](#option-b-add-authentication-to-a-loaded-tool)
|
||||
- [Option C: Add Authentication While Loading Tools](#option-c-add-authentication-while-loading-tools)
|
||||
- [Complete Authentication Example](#complete-authentication-example)
|
||||
- [Binding Parameter Values](#binding-parameter-values)
|
||||
- [Why Bind Parameters?](#why-bind-parameters)
|
||||
- [Option A: Add Default Bound Parameters to a Client](#option-a-add-default-bound-parameters-to-a-client)
|
||||
- [Option B: Binding Parameters to a Loaded Tool](#option-b-binding-parameters-to-a-loaded-tool)
|
||||
- [Option C: Binding Parameters While Loading Tools](#option-c-binding-parameters-while-loading-tools)
|
||||
- [Binding Dynamic Values](#binding-dynamic-values)
|
||||
- [Using with Orchestration Frameworks](#using-with-orchestration-frameworks)
|
||||
- [Contributing](#contributing)
|
||||
- [License](#license)
|
||||
- [Support](#support)
|
||||
|
||||
<!-- /TOC -->
|
||||
|
||||
## Installation
|
||||
|
||||
```bash
|
||||
go get github.com/googleapis/mcp-toolbox-sdk-go
|
||||
```
|
||||
This SDK is supported on Go version 1.24.4 and higher.
|
||||
|
||||
> [!NOTE]
|
||||
>
|
||||
> - While the SDK itself is synchronous, you can execute its functions within goroutines to achieve asynchronous behavior.
|
||||
|
||||
|
||||
## Quickstart
|
||||
|
||||
Here's a minimal example to get you started. Ensure your Toolbox service is
|
||||
running and accessible.
|
||||
|
||||
```go
|
||||
package main
|
||||
|
||||
import (
|
||||
"context"
|
||||
"fmt"
|
||||
"github.com/googleapis/mcp-toolbox-sdk-go/core"
|
||||
)
|
||||
|
||||
func quickstart() string {
|
||||
ctx := context.Background()
|
||||
inputs := map[string]any{"location": "London"}
|
||||
client, err := core.NewToolboxClient("http://localhost:5000")
|
||||
if err != nil {
|
||||
return fmt.Sprintln("Could not start Toolbox Client", err)
|
||||
}
|
||||
tool, err := client.LoadTool("get_weather", ctx)
|
||||
if err != nil {
|
||||
return fmt.Sprintln("Could not load Toolbox Tool", err)
|
||||
}
|
||||
result, err := tool.Invoke(ctx, inputs)
|
||||
if err != nil {
|
||||
return fmt.Sprintln("Could not invoke tool", err)
|
||||
}
|
||||
return fmt.Sprintln(result)
|
||||
}
|
||||
|
||||
func main() {
|
||||
fmt.Println(quickstart())
|
||||
}
|
||||
```
|
||||
|
||||
## Usage
|
||||
|
||||
Import and initialize a Toolbox client, pointing it to the URL of your running
|
||||
Toolbox service.
|
||||
|
||||
```go
|
||||
import "github.com/googleapis/mcp-toolbox-sdk-go/core"
|
||||
|
||||
client, err := core.NewToolboxClient("http://localhost:5000")
|
||||
```
|
||||
|
||||
All interactions for loading and invoking tools happen through this client.
|
||||
|
||||
> [!NOTE]
|
||||
> For advanced use cases, you can provide an external custom `http.Client`
|
||||
> during initialization (e.g., `core.NewToolboxClient(URL, core.WithHTTPClient(myClient)`). If you
|
||||
> provide your own session, you are responsible for managing its lifecycle;
|
||||
> `ToolboxClient` *will not* close it.
|
||||
|
||||
> [!IMPORTANT]
|
||||
> Closing the `ToolboxClient` also closes the underlying network session shared by
|
||||
> all tools loaded from that client. As a result, any tool instances you have
|
||||
> loaded will cease to function and will raise an error if you attempt to invoke
|
||||
> them after the client is closed.
|
||||
|
||||
## Transport Protocols
|
||||
|
||||
The SDK supports multiple transport protocols for communicating with the Toolbox server. By default, the client uses the latest supported version of the **Model Context Protocol (MCP)**.
|
||||
|
||||
You can explicitly select a protocol using the `core.WithProtocol` option during client initialization. This is useful if you need to use the native Toolbox HTTP protocol or pin the client to a specific legacy version of MCP.
|
||||
|
||||
> [!NOTE]
|
||||
> * **Native Toolbox Transport**: This uses the service's native **REST over HTTP** API.
|
||||
> * **MCP Transports**: These options use the **Model Context Protocol over HTTP**.
|
||||
|
||||
### Supported Protocols
|
||||
|
||||
| Constant | Description |
|
||||
| :--- | :--- |
|
||||
| `core.MCP` | **(Default)** Alias for the latest supported MCP version (currently `v2025-06-18`). |
|
||||
| `core.Toolbox` | The native Toolbox HTTP protocol. |
|
||||
| `core.MCPv20250618` | MCP Protocol version 2025-06-18. |
|
||||
| `core.MCPv20250326` | MCP Protocol version 2025-03-26. |
|
||||
| `core.MCPv20241105` | MCP Protocol version 2024-11-05. |
|
||||
|
||||
### Example
|
||||
|
||||
If you wish to use the native Toolbox protocol:
|
||||
|
||||
```go
|
||||
import "github.com/googleapis/mcp-toolbox-sdk-go/core"
|
||||
|
||||
client, err := core.NewToolboxClient(
|
||||
"http://localhost:5000",
|
||||
core.WithProtocol(core.Toolbox),
|
||||
)
|
||||
```
|
||||
If you want to pin the MCP Version 2025-03-26:
|
||||
|
||||
```go
|
||||
import "github.com/googleapis/mcp-toolbox-sdk-go/core"
|
||||
|
||||
client, err := core.NewToolboxClient(
|
||||
"http://localhost:5000",
|
||||
core.WithProtocol(core.MCPv20250326),
|
||||
)
|
||||
```
|
||||
|
||||
## Loading Tools
|
||||
|
||||
You can load tools individually or in groups (toolsets) as defined in your
|
||||
Toolbox service configuration. Loading a toolset is convenient when working with
|
||||
multiple related functions, while loading a single tool offers more granular
|
||||
control.
|
||||
|
||||
### Load a toolset
|
||||
|
||||
A toolset is a collection of related tools. You can load all tools in a toolset
|
||||
or a specific one:
|
||||
|
||||
```go
|
||||
// Load default toolset by providing an empty string as the name
|
||||
tools, err := client.LoadToolset("", ctx)
|
||||
|
||||
// Load a specific toolset
|
||||
tools, err := client.LoadToolset("my-toolset", ctx)
|
||||
```
|
||||
|
||||
|
||||
### Load a single tool
|
||||
|
||||
Loads a specific tool by its unique name. This provides fine-grained control.
|
||||
|
||||
```go
|
||||
tool, err = client.LoadTool("my-tool", ctx)
|
||||
```
|
||||
|
||||
## Invoking Tools
|
||||
|
||||
Once loaded, tools behave like Go structs. You invoke them using `Invoke` method
|
||||
by passing arguments corresponding to the parameters defined in the tool's
|
||||
configuration within the Toolbox service.
|
||||
|
||||
```go
|
||||
tool, err = client.LoadTool("my-tool", ctx)
|
||||
inputs := map[string]any{"location": "London"}
|
||||
result, err := tool.Invoke(ctx, inputs)
|
||||
```
|
||||
|
||||
> [!TIP]
|
||||
> For a more comprehensive guide on setting up the Toolbox service itself, which
|
||||
> you'll need running to use this SDK, please refer to the [Toolbox Quickstart
|
||||
> Guide](https://googleapis.github.io/genai-toolbox/getting-started/local_quickstart).
|
||||
|
||||
## Client to Server Authentication
|
||||
|
||||
This section describes how to authenticate the ToolboxClient itself when
|
||||
connecting to a Toolbox server instance that requires authentication. This is
|
||||
crucial for securing your Toolbox server endpoint, especially when deployed on
|
||||
platforms like Cloud Run, GKE, or any environment where unauthenticated access is restricted.
|
||||
|
||||
This client-to-server authentication ensures that the Toolbox server can verify
|
||||
the identity of the client making the request before any tool is loaded or
|
||||
called. It is different from [Authenticating Tools](#authenticating-tools),
|
||||
which deals with providing credentials for specific tools within an already
|
||||
connected Toolbox session.
|
||||
|
||||
### When is Client-to-Server Authentication Needed?
|
||||
|
||||
You'll need this type of authentication if your Toolbox server is configured to
|
||||
deny unauthenticated requests. For example:
|
||||
|
||||
- Your Toolbox server is deployed on Cloud Run and configured to "Require authentication."
|
||||
- Your server is behind an Identity-Aware Proxy (IAP) or a similar
|
||||
authentication layer.
|
||||
- You have custom authentication middleware on your self-hosted Toolbox server.
|
||||
|
||||
Without proper client authentication in these scenarios, attempts to connect or
|
||||
make calls (like `LoadTool`) will likely fail with `Unauthorized` errors.
|
||||
|
||||
### How it works
|
||||
|
||||
The `ToolboxClient` allows you to specify TokenSources that dynamically generate HTTP headers for
|
||||
every request sent to the Toolbox server. The most common use case is to add an
|
||||
Authorization header with a bearer token (e.g., a Google ID token).
|
||||
|
||||
These header-generating functions are called just before each request, ensuring
|
||||
that fresh credentials or header values can be used.
|
||||
|
||||
### Configuration
|
||||
|
||||
You can configure these dynamic headers as seen below:
|
||||
|
||||
|
||||
```go
|
||||
import "github.com/googleapis/mcp-toolbox-sdk-go/core"
|
||||
|
||||
tokenProvider := func() string {
|
||||
return "header3_value"
|
||||
}
|
||||
|
||||
staticTokenSource := oauth2.StaticTokenSource(&oauth2.Token{AccessToken: "header2_value"})
|
||||
dynamicTokenSource := core.NewCustomTokenSource(tokenProvider)
|
||||
|
||||
client, err := core.NewToolboxClient(
|
||||
"toolbox-url",
|
||||
core.WithClientHeaderString("header1", "header1_value"),
|
||||
core.WithClientHeaderTokenSource("header2", staticTokenSource),
|
||||
core.WithClientHeaderTokenSource("header3", dynamicTokenSource),
|
||||
)
|
||||
```
|
||||
|
||||
### Authenticating with Google Cloud Servers
|
||||
|
||||
For Toolbox servers hosted on Google Cloud (e.g., Cloud Run) and requiring
|
||||
`Google ID token` authentication, the helper module
|
||||
[auth_methods](https://github.com/googleapis/mcp-toolbox-sdk-go/blob/main/core/auth.go) provides utility functions.
|
||||
|
||||
### Step by Step Guide for Cloud Run
|
||||
|
||||
1. **Configure Permissions**: [Grant](https://cloud.google.com/run/docs/securing/managing-access#service-add-principals) the `roles/run.invoker` IAM role on the Cloud
|
||||
Run service to the principal. This could be your `user account email` or a
|
||||
`service account`.
|
||||
2. **Configure Credentials**
|
||||
- Local Development: Set up
|
||||
[ADC](https://cloud.google.com/docs/authentication/set-up-adc-local-dev-environment).
|
||||
- Google Cloud Environments: When running within Google Cloud (e.g., Compute
|
||||
Engine, GKE, another Cloud Run service, Cloud Functions), ADC is typically
|
||||
configured automatically, using the environment's default service account.
|
||||
3. **Connect to the Toolbox Server**
|
||||
```go
|
||||
import "github.com/googleapis/mcp-toolbox-sdk-go/core"
|
||||
import "context"
|
||||
|
||||
ctx := context.Background()
|
||||
|
||||
token, err := core.GetGoogleIDToken(ctx, URL)
|
||||
|
||||
client, err := core.NewToolboxClient(
|
||||
URL,
|
||||
core.WithClientHeaderString("Authorization", token),
|
||||
)
|
||||
|
||||
// Now, you can use the client as usual.
|
||||
```
|
||||
|
||||
## Authenticating Tools
|
||||
|
||||
> [!WARNING]
|
||||
> **Always use HTTPS** to connect your application with the Toolbox service,
|
||||
> especially in **production environments** or whenever the communication
|
||||
> involves **sensitive data** (including scenarios where tools require
|
||||
> authentication tokens). Using plain HTTP lacks encryption and exposes your
|
||||
> application and data to significant security risks, such as eavesdropping and
|
||||
> tampering.
|
||||
|
||||
Tools can be configured within the Toolbox service to require authentication,
|
||||
ensuring only authorized users or applications can invoke them, especially when
|
||||
accessing sensitive data.
|
||||
|
||||
### When is Authentication Needed?
|
||||
|
||||
Authentication is configured per-tool within the Toolbox service itself. If a
|
||||
tool you intend to use is marked as requiring authentication in the service, you
|
||||
must configure the SDK client to provide the necessary credentials (currently
|
||||
Oauth2 tokens) when invoking that specific tool.
|
||||
|
||||
### Supported Authentication Mechanisms
|
||||
|
||||
The Toolbox service enables secure tool usage through **Authenticated Parameters**.
|
||||
For detailed information on how these mechanisms work within the Toolbox service and how to configure them, please refer to [Toolbox Service Documentation - Authenticated Parameters](https://googleapis.github.io/genai-toolbox/resources/tools/#authenticated-parameters).
|
||||
|
||||
### Step 1: Configure Tools in Toolbox Service
|
||||
|
||||
First, ensure the target tool(s) are configured correctly in the Toolbox service
|
||||
to require authentication. Refer to the [Toolbox Service Documentation -
|
||||
Authenticated
|
||||
Parameters](https://googleapis.github.io/genai-toolbox/resources/tools/#authenticated-parameters)
|
||||
for instructions.
|
||||
|
||||
### Step 2: Configure SDK Client
|
||||
|
||||
Your application needs a way to obtain the required Oauth2 token for the
|
||||
authenticated user. The SDK requires you to provide a function capable of
|
||||
retrieving this token *when the tool is invoked*.
|
||||
|
||||
#### Provide an ID Token Retriever Function
|
||||
|
||||
You must provide the SDK with a function that returns the
|
||||
necessary token when called. The implementation depends on your application's
|
||||
authentication flow (e.g., retrieving a stored token, initiating an OAuth flow).
|
||||
|
||||
> [!IMPORTANT]
|
||||
> The name used when registering the getter function with the SDK (e.g.,
|
||||
> `"my_api_token"`) must exactly match the `name` of the corresponding
|
||||
> `authServices` defined in the tool's configuration within the Toolbox service.
|
||||
|
||||
```go
|
||||
func getAuthToken() string {
|
||||
// ... Logic to retrieve ID token (e.g., from local storage, OAuth flow)
|
||||
// This example just returns a placeholder. Replace with your actual token retrieval.
|
||||
return "YOUR_ID_TOKEN" // Placeholder
|
||||
}
|
||||
```
|
||||
|
||||
> [!TIP]
|
||||
> Your token retriever function is invoked every time an authenticated parameter
|
||||
> requires a token for a tool call. Consider implementing caching logic within
|
||||
> this function to avoid redundant token fetching or generation, especially for
|
||||
> tokens with longer validity periods or if the retrieval process is
|
||||
> resource-intensive.
|
||||
|
||||
#### Option A: Add Default Authentication to a Client
|
||||
|
||||
You can add default tool level authentication to a client.
|
||||
Every tool / toolset loaded by the client will contain the auth token.
|
||||
|
||||
```go
|
||||
|
||||
ctx := context.Background()
|
||||
|
||||
client, err := core.NewToolboxClient("http://127.0.0.1:5000",
|
||||
core.WithDefaultToolOptions(
|
||||
core.WithAuthTokenString("my-auth-1", "auth-value"),
|
||||
),
|
||||
)
|
||||
|
||||
AuthTool, err := client.LoadTool("my-tool", ctx)
|
||||
```
|
||||
|
||||
#### Option B: Add Authentication to a Loaded Tool
|
||||
|
||||
You can add the token retriever function to a tool object *after* it has been
|
||||
loaded. This modifies the specific tool instance.
|
||||
|
||||
```go
|
||||
|
||||
ctx := context.Background()
|
||||
|
||||
client, err := core.NewToolboxClient("http://127.0.0.1:5000")
|
||||
|
||||
tool, err := client.LoadTool("my-tool", ctx)
|
||||
|
||||
AuthTool, err := tool.ToolFrom(
|
||||
core.WithAuthTokenSource("my-auth", headerTokenSource),
|
||||
core.WithAuthTokenString("my-auth-1", "value"),
|
||||
)
|
||||
```
|
||||
|
||||
#### Option C: Add Authentication While Loading Tools
|
||||
|
||||
You can provide the token retriever(s) directly during the `LoadTool` or
|
||||
`LoadToolset` calls. This applies the authentication configuration only to the
|
||||
tools loaded in that specific call, without modifying the original tool objects
|
||||
if they were loaded previously.
|
||||
|
||||
```go
|
||||
AuthTool, err := client.LoadTool("my-tool", ctx, core.WithAuthTokenString("my-auth-1", "value"))
|
||||
|
||||
// or
|
||||
|
||||
AuthTools, err := client.LoadToolset(
|
||||
"my-toolset",
|
||||
ctx,
|
||||
core.WithAuthTokenString("my-auth-1", "value"),
|
||||
)
|
||||
```
|
||||
|
||||
> [!NOTE]
|
||||
> Adding auth tokens during loading only affect the tools loaded within that
|
||||
> call.
|
||||
|
||||
### Complete Authentication Example
|
||||
|
||||
```go
|
||||
import "github.com/googleapis/mcp-toolbox-sdk-go/core"
|
||||
import "fmt"
|
||||
|
||||
func getAuthToken() string {
|
||||
// ... Logic to retrieve ID token (e.g., from local storage, OAuth flow)
|
||||
// This example just returns a placeholder. Replace with your actual token retrieval.
|
||||
return "YOUR_ID_TOKEN" // Placeholder
|
||||
}
|
||||
|
||||
func main() {
|
||||
ctx := context.Background()
|
||||
inputs := map[string]any{"input": "some input"}
|
||||
|
||||
dynamicTokenSource := core.NewCustomTokenSource(getAuthToken)
|
||||
|
||||
client, err := core.NewToolboxClient("http://127.0.0.1:5000")
|
||||
tool, err := client.LoadTool("my-tool", ctx)
|
||||
AuthTool, err := tool.ToolFrom(core.WithAuthTokenSource("my_auth", dynamicTokenSource))
|
||||
|
||||
result, err := AuthTool.Invoke(ctx, inputs)
|
||||
|
||||
fmt.Println(result)
|
||||
}
|
||||
```
|
||||
|
||||
> [!NOTE]
|
||||
> An auth token getter for a specific name (e.g., "GOOGLE_ID") will replace any
|
||||
> client header with the same name followed by "_token" (e.g.,
|
||||
> "GOOGLE_ID_token").
|
||||
|
||||
## Binding Parameter Values
|
||||
|
||||
The SDK allows you to pre-set, or "bind", values for specific tool parameters
|
||||
before the tool is invoked or even passed to an LLM. These bound values are
|
||||
fixed and will not be requested or modified by the LLM during tool use.
|
||||
|
||||
### Why Bind Parameters?
|
||||
|
||||
- **Protecting sensitive information:** API keys, secrets, etc.
|
||||
- **Enforcing consistency:** Ensuring specific values for certain parameters.
|
||||
- **Pre-filling known data:** Providing defaults or context.
|
||||
|
||||
> [!IMPORTANT]
|
||||
> The parameter names used for binding (e.g., `"api_key"`) must exactly match the
|
||||
> parameter names defined in the tool's configuration within the Toolbox
|
||||
> service.
|
||||
|
||||
> [!NOTE]
|
||||
> You do not need to modify the tool's configuration in the Toolbox service to
|
||||
> bind parameter values using the SDK.
|
||||
|
||||
#### Option A: Add Default Bound Parameters to a Client
|
||||
|
||||
You can add default tool level bound parameters to a client. Every tool / toolset
|
||||
loaded by the client will have the bound parameter.
|
||||
|
||||
```go
|
||||
|
||||
ctx := context.Background()
|
||||
|
||||
client, err := core.NewToolboxClient("http://127.0.0.1:5000",
|
||||
core.WithDefaultToolOptions(
|
||||
core.WithBindParamString("param1", "value"),
|
||||
),
|
||||
)
|
||||
|
||||
boundTool, err := client.LoadTool("my-tool", ctx)
|
||||
```
|
||||
|
||||
### Option B: Binding Parameters to a Loaded Tool
|
||||
|
||||
Bind values to a tool object *after* it has been loaded. This modifies the
|
||||
specific tool instance.
|
||||
|
||||
```go
|
||||
client, err := core.NewToolboxClient("http://127.0.0.1:5000")
|
||||
|
||||
tool, err := client.LoadTool("my-tool", ctx)
|
||||
|
||||
boundTool, err := tool.ToolFrom(
|
||||
core.WithBindParamString("param1", "value"),
|
||||
core.WithBindParamString("param2", "value")
|
||||
)
|
||||
```
|
||||
|
||||
### Option C: Binding Parameters While Loading Tools
|
||||
|
||||
Specify bound parameters directly when loading tools. This applies the binding
|
||||
only to the tools loaded in that specific call.
|
||||
|
||||
```go
|
||||
boundTool, err := client.LoadTool("my-tool", ctx, core.WithBindParamString("param", "value"))
|
||||
|
||||
// OR
|
||||
|
||||
boundTool, err := client.LoadToolset("", ctx, core.WithBindParamString("param", "value"))
|
||||
```
|
||||
|
||||
> [!NOTE]
|
||||
> Bound values during loading only affect the tools loaded in that call.
|
||||
|
||||
### Binding Dynamic Values
|
||||
|
||||
Instead of a static value, you can bind a parameter to a synchronous or
|
||||
asynchronous function. This function will be called *each time* the tool is
|
||||
invoked to dynamically determine the parameter's value at runtime.
|
||||
Functions with the return type (data_type, error) can be provided.
|
||||
|
||||
```go
|
||||
getDynamicValue := func() (string, error) { return "req-123", nil }
|
||||
|
||||
dynamicBoundTool, err := tool.ToolFrom(core.WithBindParamStringFunc("param", getDynamicValue))
|
||||
```
|
||||
|
||||
> [!IMPORTANT]
|
||||
> You don't need to modify tool configurations to bind parameter values.
|
||||
|
||||
|
||||
# Using with Orchestration Frameworks
|
||||
|
||||
To see how the MCP Toolbox Go SDK works with orchestration frameworks, check out the end-to-end examples in the [/samples/](https://github.com/googleapis/mcp-toolbox-sdk-go/tree/main/core/samples) folder.
|
||||
|
||||
Use the [tbgenkit package](https://github.com/googleapis/mcp-toolbox-sdk-go/tree/main/tbgenkit) to convert Toolbox Tools into Genkit compatible tools.
|
||||
|
||||
# Contributing
|
||||
|
||||
Contributions are welcome! Please refer to the [DEVELOPER.md](https://github.com/googleapis/mcp-toolbox-sdk-go/blob/main/DEVELOPER.md)
|
||||
file for guidelines on how to set up a development environment and run tests.
|
||||
|
||||
# License
|
||||
|
||||
This project is licensed under the Apache License 2.0. See the
|
||||
[LICENSE](https://github.com/googleapis/mcp-toolbox-sdk-go/blob/main/LICENSE) file for details.
|
||||
|
||||
# Support
|
||||
|
||||
If you encounter issues or have questions, check the existing [GitHub Issues](https://github.com/googleapis/genai-toolbox/issues) for the main Toolbox project.
|
||||
|
||||
## Samples for Reference
|
||||
|
||||
These samples demonstrate how to integrate the MCP Toolbox Go Core SDK with popular orchestration frameworks.
|
||||
|
||||
{{< tabpane persist=header >}}
|
||||
|
||||
{{< tab header="Google GenAI" lang="go" >}}
|
||||
// This sample demonstrates integration with the standard Google GenAI framework.
|
||||
package main
|
||||
|
||||
import (
|
||||
"context"
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"log"
|
||||
"os"
|
||||
|
||||
"github.com/googleapis/mcp-toolbox-sdk-go/core"
|
||||
"google.golang.org/genai"
|
||||
)
|
||||
|
||||
// ConvertToGenaiTool translates a ToolboxTool into the genai.FunctionDeclaration format.
|
||||
func ConvertToGenaiTool(toolboxTool *core.ToolboxTool) *genai.Tool {
|
||||
|
||||
inputschema, err := toolboxTool.InputSchema()
|
||||
if err != nil {
|
||||
return &genai.Tool{}
|
||||
}
|
||||
|
||||
var schema *genai.Schema
|
||||
_ = json.Unmarshal(inputschema, &schema)
|
||||
// First, create the function declaration.
|
||||
funcDeclaration := &genai.FunctionDeclaration{
|
||||
Name: toolboxTool.Name(),
|
||||
Description: toolboxTool.Description(),
|
||||
Parameters: schema,
|
||||
}
|
||||
|
||||
// Then, wrap the function declaration in a genai.Tool struct.
|
||||
return &genai.Tool{
|
||||
FunctionDeclarations: []*genai.FunctionDeclaration{funcDeclaration},
|
||||
}
|
||||
}
|
||||
|
||||
// printResponse extracts and prints the relevant parts of the model's response.
|
||||
func printResponse(resp *genai.GenerateContentResponse) {
|
||||
for _, cand := range resp.Candidates {
|
||||
if cand.Content != nil {
|
||||
for _, part := range cand.Content.Parts {
|
||||
fmt.Println(part.Text)
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func main() {
|
||||
// Setup
|
||||
ctx := context.Background()
|
||||
apiKey := os.Getenv("GOOGLE_API_KEY")
|
||||
toolboxURL := "http://localhost:5000"
|
||||
|
||||
// Initialize the Google GenAI client using the explicit ClientConfig.
|
||||
client, err := genai.NewClient(ctx, &genai.ClientConfig{
|
||||
APIKey: apiKey,
|
||||
})
|
||||
if err != nil {
|
||||
log.Fatalf("Failed to create Google GenAI client: %v", err)
|
||||
}
|
||||
|
||||
// Initialize the MCP Toolbox client.
|
||||
toolboxClient, err := core.NewToolboxClient(toolboxURL)
|
||||
if err != nil {
|
||||
log.Fatalf("Failed to create Toolbox client: %v", err)
|
||||
}
|
||||
|
||||
// Load the tools using the MCP Toolbox SDK.
|
||||
tools, err := toolboxClient.LoadToolset("my-toolset", ctx)
|
||||
if err != nil {
|
||||
log.Fatalf("Failed to load tools: %v\nMake sure your Toolbox server is running and the tool is configured.", err)
|
||||
}
|
||||
|
||||
genAITools := make([]*genai.Tool, len(tools))
|
||||
toolsMap := make(map[string]*core.ToolboxTool, len(tools))
|
||||
|
||||
for i, tool := range tools {
|
||||
// Convert the tools into usable format
|
||||
genAITools[i] = ConvertToGenaiTool(tool)
|
||||
// Add tool to a map for lookup later
|
||||
toolsMap[tool.Name()] = tool
|
||||
}
|
||||
|
||||
// Set up the generative model with the available tool.
|
||||
modelName := "gemini-2.0-flash"
|
||||
|
||||
query := "Find hotels in Basel with Basel in it's name and share the names with me"
|
||||
|
||||
// Create the initial content prompt for the model.
|
||||
contents := []*genai.Content{
|
||||
genai.NewContentFromText(query, genai.RoleUser),
|
||||
}
|
||||
config := &genai.GenerateContentConfig{
|
||||
Tools: genAITools,
|
||||
ToolConfig: &genai.ToolConfig{
|
||||
FunctionCallingConfig: &genai.FunctionCallingConfig{
|
||||
Mode: genai.FunctionCallingConfigModeAny,
|
||||
},
|
||||
},
|
||||
}
|
||||
genContentResp, _ := client.Models.GenerateContent(ctx, modelName, contents, config)
|
||||
|
||||
printResponse(genContentResp)
|
||||
|
||||
functionCalls := genContentResp.FunctionCalls()
|
||||
if len(functionCalls) == 0 {
|
||||
log.Println("No function call returned by the AI. The model likely answered directly.")
|
||||
return
|
||||
}
|
||||
|
||||
// Process the first function call (the example assumes one for simplicity).
|
||||
fc := functionCalls[0]
|
||||
log.Printf("--- Gemini requested function call: %s ---\n", fc.Name)
|
||||
log.Printf("--- Arguments: %+v ---\n", fc.Args)
|
||||
|
||||
var toolResultString string
|
||||
|
||||
if fc.Name == "search-hotels-by-name" {
|
||||
tool := toolsMap["search-hotels-by-name"]
|
||||
toolResult, err := tool.Invoke(ctx, fc.Args)
|
||||
toolResultString = fmt.Sprintf("%v", toolResult)
|
||||
if err != nil {
|
||||
log.Fatalf("Failed to execute tool '%s': %v", fc.Name, err)
|
||||
}
|
||||
|
||||
} else {
|
||||
log.Println("LLM did not request our tool")
|
||||
}
|
||||
resultContents := []*genai.Content{
|
||||
genai.NewContentFromText("The tool returned this result, share it with the user based of their previous querys"+toolResultString, genai.RoleUser),
|
||||
}
|
||||
finalResponse, err := client.Models.GenerateContent(ctx, modelName, resultContents, &genai.GenerateContentConfig{})
|
||||
if err != nil {
|
||||
log.Fatalf("Error calling GenerateContent (with function result): %v", err)
|
||||
}
|
||||
log.Println("=== Final Response from Model (after processing function result) ===")
|
||||
printResponse(finalResponse)
|
||||
|
||||
}
|
||||
{{< /tab >}}
|
||||
|
||||
{{< tab header="LangChain Go" lang="go" >}}
|
||||
// This sample demonstrates how to use Toolbox tools as function definitions in LangChain Go.
|
||||
package main
|
||||
|
||||
import (
|
||||
"context"
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"log"
|
||||
"os"
|
||||
|
||||
"github.com/googleapis/mcp-toolbox-sdk-go/core"
|
||||
"github.com/tmc/langchaingo/llms"
|
||||
"github.com/tmc/langchaingo/llms/googleai"
|
||||
)
|
||||
|
||||
// ConvertToLangchainTool converts a generic core.ToolboxTool into a LangChainGo llms.Tool.
|
||||
func ConvertToLangchainTool(toolboxTool *core.ToolboxTool) llms.Tool {
|
||||
|
||||
// Fetch the tool's input schema
|
||||
inputschema, err := toolboxTool.InputSchema()
|
||||
if err != nil {
|
||||
return llms.Tool{}
|
||||
}
|
||||
|
||||
var paramsSchema map[string]any
|
||||
_ = json.Unmarshal(inputschema, ¶msSchema)
|
||||
|
||||
// Convert into LangChain's llms.Tool
|
||||
return llms.Tool{
|
||||
Type: "function",
|
||||
Function: &llms.FunctionDefinition{
|
||||
Name: toolboxTool.Name(),
|
||||
Description: toolboxTool.Description(),
|
||||
Parameters: paramsSchema,
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
func main() {
|
||||
genaiKey := os.Getenv("GOOGLE_API_KEY")
|
||||
toolboxURL := "http://localhost:5000"
|
||||
ctx := context.Background()
|
||||
|
||||
// Initialize the Google AI client (LLM).
|
||||
llm, err := googleai.New(ctx, googleai.WithAPIKey(genaiKey), googleai.WithDefaultModel("gemini-1.5-flash"))
|
||||
if err != nil {
|
||||
log.Fatalf("Failed to create Google AI client: %v", err)
|
||||
}
|
||||
|
||||
// Initialize the MCP Toolbox client.
|
||||
toolboxClient, err := core.NewToolboxClient(toolboxURL)
|
||||
if err != nil {
|
||||
log.Fatalf("Failed to create Toolbox client: %v", err)
|
||||
}
|
||||
|
||||
// Load the tools using the MCP Toolbox SDK.
|
||||
tools, err := toolboxClient.LoadToolset("my-toolset", ctx)
|
||||
if err != nil {
|
||||
log.Fatalf("Failed to load tools: %v\nMake sure your Toolbox server is running and the tool is configured.", err)
|
||||
}
|
||||
|
||||
toolsMap := make(map[string]*core.ToolboxTool, len(tools))
|
||||
|
||||
langchainTools := make([]llms.Tool, len(tools))
|
||||
for i, tool := range tools {
|
||||
// Convert the loaded ToolboxTools into the format LangChainGo requires.
|
||||
langchainTools[i] = ConvertToLangchainTool(tool)
|
||||
// Add tool to a map for lookup later
|
||||
toolsMap[tool.Name()] = tool
|
||||
}
|
||||
|
||||
// Start the conversation history.
|
||||
messageHistory := []llms.MessageContent{
|
||||
llms.TextParts(llms.ChatMessageTypeHuman, "Find hotels in Basel with Basel in it's name."),
|
||||
}
|
||||
|
||||
// Make the first call to the LLM, making it aware of the tool.
|
||||
resp, err := llm.GenerateContent(ctx, messageHistory, llms.WithTools(langchainTools))
|
||||
if err != nil {
|
||||
log.Fatalf("LLM call failed: %v", err)
|
||||
}
|
||||
|
||||
// Add the model's response (which should be a tool call) to the history.
|
||||
respChoice := resp.Choices[0]
|
||||
assistantResponse := llms.TextParts(llms.ChatMessageTypeAI, respChoice.Content)
|
||||
for _, tc := range respChoice.ToolCalls {
|
||||
assistantResponse.Parts = append(assistantResponse.Parts, tc)
|
||||
}
|
||||
messageHistory = append(messageHistory, assistantResponse)
|
||||
|
||||
// Process each tool call requested by the model.
|
||||
for _, tc := range respChoice.ToolCalls {
|
||||
toolName := tc.FunctionCall.Name
|
||||
|
||||
switch tc.FunctionCall.Name {
|
||||
case "search-hotels-by-name":
|
||||
var args map[string]any
|
||||
if err := json.Unmarshal([]byte(tc.FunctionCall.Arguments), &args); err != nil {
|
||||
log.Fatalf("Failed to unmarshal arguments for tool '%s': %v", toolName, err)
|
||||
}
|
||||
tool := toolsMap["search-hotels-by-name"]
|
||||
toolResult, err := tool.Invoke(ctx, args)
|
||||
if err != nil {
|
||||
log.Fatalf("Failed to execute tool '%s': %v", toolName, err)
|
||||
}
|
||||
|
||||
// Create the tool call response message and add it to the history.
|
||||
toolResponse := llms.MessageContent{
|
||||
Role: llms.ChatMessageTypeTool,
|
||||
Parts: []llms.ContentPart{
|
||||
llms.ToolCallResponse{
|
||||
Name: toolName,
|
||||
Content: fmt.Sprintf("%v", toolResult),
|
||||
},
|
||||
},
|
||||
}
|
||||
messageHistory = append(messageHistory, toolResponse)
|
||||
default:
|
||||
log.Fatalf("got unexpected function call: %v", tc.FunctionCall.Name)
|
||||
}
|
||||
}
|
||||
|
||||
// Final LLM Call for Natural Language Response
|
||||
log.Println("Sending tool response back to LLM for a final answer...")
|
||||
|
||||
// Call the LLM again with the updated history, which now includes the tool's result.
|
||||
finalResp, err := llm.GenerateContent(ctx, messageHistory)
|
||||
if err != nil {
|
||||
log.Fatalf("Final LLM call failed: %v", err)
|
||||
}
|
||||
|
||||
// Display the Result
|
||||
fmt.Println("\n======================================")
|
||||
fmt.Println("Final Response from LLM:")
|
||||
fmt.Println(finalResp.Choices[0].Content)
|
||||
fmt.Println("======================================")
|
||||
}
|
||||
{{< /tab >}}
|
||||
|
||||
{{< tab header="OpenAI Go" lang="go" >}}
|
||||
// This sample demonstrates integration with the OpenAI Go client.
|
||||
package main
|
||||
|
||||
import (
|
||||
"context"
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"log"
|
||||
|
||||
"github.com/googleapis/mcp-toolbox-sdk-go/core"
|
||||
openai "github.com/openai/openai-go"
|
||||
)
|
||||
|
||||
// ConvertToOpenAITool converts a ToolboxTool into the go-openai library's Tool format.
|
||||
func ConvertToOpenAITool(toolboxTool *core.ToolboxTool) openai.ChatCompletionToolParam {
|
||||
// Get the input schema
|
||||
jsonSchemaBytes, err := toolboxTool.InputSchema()
|
||||
if err != nil {
|
||||
return openai.ChatCompletionToolParam{}
|
||||
}
|
||||
|
||||
// Unmarshal the JSON bytes into FunctionParameters
|
||||
var paramsSchema openai.FunctionParameters
|
||||
if err := json.Unmarshal(jsonSchemaBytes, ¶msSchema); err != nil {
|
||||
return openai.ChatCompletionToolParam{}
|
||||
}
|
||||
|
||||
// Create and return the final tool parameter struct.
|
||||
return openai.ChatCompletionToolParam{
|
||||
Function: openai.FunctionDefinitionParam{
|
||||
Name: toolboxTool.Name(),
|
||||
Description: openai.String(toolboxTool.Description()),
|
||||
Parameters: paramsSchema,
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
func main() {
|
||||
// Setup
|
||||
ctx := context.Background()
|
||||
toolboxURL := "http://localhost:5000"
|
||||
openAIClient := openai.NewClient()
|
||||
|
||||
// Initialize the MCP Toolbox client.
|
||||
toolboxClient, err := core.NewToolboxClient(toolboxURL)
|
||||
if err != nil {
|
||||
log.Fatalf("Failed to create Toolbox client: %v", err)
|
||||
}
|
||||
|
||||
// Load the tools using the MCP Toolbox SDK.
|
||||
tools, err := toolboxClient.LoadToolset("my-toolset", ctx)
|
||||
if err != nil {
|
||||
log.Fatalf("Failed to load tool : %v\nMake sure your Toolbox server is running and the tool is configured.", err)
|
||||
}
|
||||
|
||||
openAITools := make([]openai.ChatCompletionToolParam, len(tools))
|
||||
toolsMap := make(map[string]*core.ToolboxTool, len(tools))
|
||||
|
||||
for i, tool := range tools {
|
||||
// Convert the Toolbox tool into the openAI FunctionDeclaration format.
|
||||
openAITools[i] = ConvertToOpenAITool(tool)
|
||||
// Add tool to a map for lookup later
|
||||
toolsMap[tool.Name()] = tool
|
||||
|
||||
}
|
||||
question := "Find hotels in Basel with Basel in it's name "
|
||||
|
||||
params := openai.ChatCompletionNewParams{
|
||||
Messages: []openai.ChatCompletionMessageParamUnion{
|
||||
openai.UserMessage(question),
|
||||
},
|
||||
Tools: openAITools,
|
||||
Seed: openai.Int(0),
|
||||
Model: openai.ChatModelGPT4o,
|
||||
}
|
||||
|
||||
// Make initial chat completion request
|
||||
completion, err := openAIClient.Chat.Completions.New(ctx, params)
|
||||
if err != nil {
|
||||
panic(err)
|
||||
}
|
||||
|
||||
toolCalls := completion.Choices[0].Message.ToolCalls
|
||||
|
||||
// Return early if there are no tool calls
|
||||
if len(toolCalls) == 0 {
|
||||
fmt.Printf("No function call")
|
||||
return
|
||||
}
|
||||
|
||||
// If there is a was a function call, continue the conversation
|
||||
params.Messages = append(params.Messages, completion.Choices[0].Message.ToParam())
|
||||
for _, toolCall := range toolCalls {
|
||||
if toolCall.Function.Name == "search-hotels-by-name" {
|
||||
// Extract the location from the function call arguments
|
||||
var args map[string]interface{}
|
||||
tool := toolsMap["search-hotels-by-name"]
|
||||
err := json.Unmarshal([]byte(toolCall.Function.Arguments), &args)
|
||||
if err != nil {
|
||||
panic(err)
|
||||
}
|
||||
|
||||
result, err := tool.Invoke(ctx, args)
|
||||
if err != nil {
|
||||
log.Fatal("Could not invoke tool", err)
|
||||
}
|
||||
|
||||
params.Messages = append(params.Messages, openai.ToolMessage(result.(string), toolCall.ID))
|
||||
}
|
||||
}
|
||||
|
||||
completion, err = openAIClient.Chat.Completions.New(ctx, params)
|
||||
if err != nil {
|
||||
panic(err)
|
||||
}
|
||||
|
||||
fmt.Println(completion.Choices[0].Message.Content)
|
||||
}
|
||||
{{< /tab >}}
|
||||
|
||||
{{< /tabpane >}}
|
||||
@@ -22,4 +22,36 @@ These Python SDKs act as clients for that service. They handle the communication
|
||||
By using these SDKs, you can easily leverage your Toolbox-managed tools directly
|
||||
within your Python applications or AI orchestration frameworks.
|
||||
|
||||
[Github](https://github.com/googleapis/mcp-toolbox-sdk-python)
|
||||
## Which Package Should I Use?
|
||||
|
||||
Choosing the right package depends on how you are building your application:
|
||||
|
||||
* [`toolbox-langchain`](langchain):
|
||||
Use this package if you are building your application using the LangChain or
|
||||
LangGraph frameworks. It provides tools that are directly compatible with the
|
||||
LangChain ecosystem (`BaseTool` interface), simplifying integration.
|
||||
* [`toolbox-llamaindex`](llamaindex):
|
||||
Use this package if you are building your application using the LlamaIndex framework.
|
||||
It provides tools that are directly compatible with the
|
||||
LlamaIndex ecosystem (`BaseTool` interface), simplifying integration.
|
||||
* [`toolbox-core`](core):
|
||||
Use this package if you are not using LangChain/LangGraph or any other
|
||||
orchestration framework, or if you need a framework-agnostic way to interact
|
||||
with Toolbox tools (e.g., for custom orchestration logic or direct use in
|
||||
Python scripts).
|
||||
|
||||
## Available Packages
|
||||
|
||||
This repository hosts the following Python packages. See the package-specific
|
||||
README for detailed installation and usage instructions:
|
||||
|
||||
| Package | Target Use Case | Integration | Path | Details (README) | PyPI Status |
|
||||
| :------ | :---------- | :---------- | :---------------------- | :---------- | :---------
|
||||
| `toolbox-core` | Framework-agnostic / Custom applications | Use directly / Custom | `packages/toolbox-core/` | 📄 [View README](https://github.com/googleapis/mcp-toolbox-sdk-python/blob/main/packages/toolbox-core/README.md) |  |
|
||||
| `toolbox-langchain` | LangChain / LangGraph applications | LangChain / LangGraph | `packages/toolbox-langchain/` | 📄 [View README](https://github.com/googleapis/mcp-toolbox-sdk-python/blob/main/packages/toolbox-langchain/README.md) |  |
|
||||
| `toolbox-llamaindex` | LlamaIndex applications | LlamaIndex | `packages/toolbox-llamaindex/` | 📄 [View README](https://github.com/googleapis/mcp-toolbox-sdk-python/blob/main/packages/toolbox-llamaindex/README.md) |  |
|
||||
|
||||
|
||||
{{< notice note >}}
|
||||
Source code for [python-sdk](https://github.com/googleapis/mcp-toolbox-sdk-python)
|
||||
{{< /notice >}}
|
||||
|
||||
386
docs/en/sdks/python-sdk/llamaindex/index.md
Normal file
386
docs/en/sdks/python-sdk/llamaindex/index.md
Normal file
@@ -0,0 +1,386 @@
|
||||
---
|
||||
title: "llamaindex"
|
||||
type: docs
|
||||
weight: 8
|
||||
description: >
|
||||
Toolbox-llamaindex SDK for connecting to the MCP Toolbox server and invoking tools programmatically.
|
||||
---
|
||||
|
||||
## Overview
|
||||
|
||||
The `toolbox-llamaindex` package provides a Python interface to the MCP Toolbox service, enabling you to load and invoke tools from your own applications.
|
||||
|
||||
## Installation
|
||||
|
||||
```bash
|
||||
pip install toolbox-llamaindex
|
||||
```
|
||||
|
||||
## Quickstart
|
||||
|
||||
Here's a minimal example to get you started using
|
||||
[LlamaIndex](https://docs.llamaindex.ai/en/stable/#getting-started):
|
||||
|
||||
```py
|
||||
import asyncio
|
||||
|
||||
from llama_index.llms.google_genai import GoogleGenAI
|
||||
from llama_index.core.agent.workflow import AgentWorkflow
|
||||
|
||||
from toolbox_llamaindex import ToolboxClient
|
||||
|
||||
async def run_agent():
|
||||
async with ToolboxClient("http://127.0.0.1:5000") as toolbox:
|
||||
tools = toolbox.load_toolset()
|
||||
|
||||
vertex_model = GoogleGenAI(
|
||||
model="gemini-2.0-flash-001",
|
||||
vertexai_config={"project": "project-id", "location": "us-central1"},
|
||||
)
|
||||
agent = AgentWorkflow.from_tools_or_functions(
|
||||
tools,
|
||||
llm=vertex_model,
|
||||
system_prompt="You are a helpful assistant.",
|
||||
)
|
||||
response = await agent.run(user_msg="Get some response from the agent.")
|
||||
print(response)
|
||||
|
||||
asyncio.run(run_agent())
|
||||
```
|
||||
|
||||
{{< notice tip >}}
|
||||
For a complete, end-to-end example including setting up the service and using an SDK, see the full tutorial: [Toolbox Quickstart Tutorial](getting-started/local_quickstart)
|
||||
{{< /notice >}}
|
||||
|
||||
## Usage
|
||||
|
||||
Import and initialize the toolbox client.
|
||||
|
||||
```py
|
||||
from toolbox_llamaindex import ToolboxClient
|
||||
|
||||
# Replace with your Toolbox service's URL
|
||||
async with ToolboxClient("http://127.0.0.1:5000") as toolbox:
|
||||
```
|
||||
|
||||
## Loading Tools
|
||||
|
||||
### Load a toolset
|
||||
|
||||
A toolset is a collection of related tools. You can load all tools in a toolset
|
||||
or a specific one:
|
||||
|
||||
```py
|
||||
# Load all tools
|
||||
tools = toolbox.load_toolset()
|
||||
|
||||
# Load a specific toolset
|
||||
tools = toolbox.load_toolset("my-toolset")
|
||||
```
|
||||
|
||||
### Load a single tool
|
||||
|
||||
```py
|
||||
tool = toolbox.load_tool("my-tool")
|
||||
```
|
||||
|
||||
Loading individual tools gives you finer-grained control over which tools are
|
||||
available to your LLM agent.
|
||||
|
||||
## Use with LlamaIndex
|
||||
|
||||
LlamaIndex's agents can dynamically choose and execute tools based on the user
|
||||
input. Include tools loaded from the Toolbox SDK in the agent's toolkit:
|
||||
|
||||
```py
|
||||
from llama_index.llms.google_genai import GoogleGenAI
|
||||
from llama_index.core.agent.workflow import AgentWorkflow
|
||||
|
||||
vertex_model = GoogleGenAI(
|
||||
model="gemini-2.0-flash-001",
|
||||
vertexai_config={"project": "project-id", "location": "us-central1"},
|
||||
)
|
||||
|
||||
# Initialize agent with tools
|
||||
agent = AgentWorkflow.from_tools_or_functions(
|
||||
tools,
|
||||
llm=vertex_model,
|
||||
system_prompt="You are a helpful assistant.",
|
||||
)
|
||||
|
||||
# Query the agent
|
||||
response = await agent.run(user_msg="Get some response from the agent.")
|
||||
print(response)
|
||||
```
|
||||
|
||||
### Maintain state
|
||||
|
||||
To maintain state for the agent, add context as follows:
|
||||
|
||||
```py
|
||||
from llama_index.core.agent.workflow import AgentWorkflow
|
||||
from llama_index.core.workflow import Context
|
||||
from llama_index.llms.google_genai import GoogleGenAI
|
||||
|
||||
vertex_model = GoogleGenAI(
|
||||
model="gemini-2.0-flash-001",
|
||||
vertexai_config={"project": "project-id", "location": "us-central1"},
|
||||
)
|
||||
agent = AgentWorkflow.from_tools_or_functions(
|
||||
tools,
|
||||
llm=vertex_model,
|
||||
system_prompt="You are a helpful assistant",
|
||||
)
|
||||
|
||||
# Save memory in agent context
|
||||
ctx = Context(agent)
|
||||
response = await agent.run(user_msg="Give me some response.", ctx=ctx)
|
||||
print(response)
|
||||
```
|
||||
|
||||
## Manual usage
|
||||
|
||||
Execute a tool manually using the `call` method:
|
||||
|
||||
```py
|
||||
result = tools[0].call(name="Alice", age=30)
|
||||
```
|
||||
|
||||
This is useful for testing tools or when you need precise control over tool
|
||||
execution outside of an agent framework.
|
||||
|
||||
## Client to Server Authentication
|
||||
|
||||
This section describes how to authenticate the ToolboxClient itself when
|
||||
connecting to a Toolbox server instance that requires authentication. This is
|
||||
crucial for securing your Toolbox server endpoint, especially when deployed on
|
||||
platforms like Cloud Run, GKE, or any environment where unauthenticated access is restricted.
|
||||
|
||||
This client-to-server authentication ensures that the Toolbox server can verify
|
||||
the identity of the client making the request before any tool is loaded or
|
||||
called. It is different from [Authenticating Tools](#authenticating-tools),
|
||||
which deals with providing credentials for specific tools within an already
|
||||
connected Toolbox session.
|
||||
|
||||
### When is Client-to-Server Authentication Needed?
|
||||
|
||||
You'll need this type of authentication if your Toolbox server is configured to
|
||||
deny unauthenticated requests. For example:
|
||||
|
||||
- Your Toolbox server is deployed on Cloud Run and configured to "Require authentication."
|
||||
- Your server is behind an Identity-Aware Proxy (IAP) or a similar
|
||||
authentication layer.
|
||||
- You have custom authentication middleware on your self-hosted Toolbox server.
|
||||
|
||||
Without proper client authentication in these scenarios, attempts to connect or
|
||||
make calls (like `load_tool`) will likely fail with `Unauthorized` errors.
|
||||
|
||||
### How it works
|
||||
|
||||
The `ToolboxClient` allows you to specify functions (or coroutines for the async
|
||||
client) that dynamically generate HTTP headers for every request sent to the
|
||||
Toolbox server. The most common use case is to add an Authorization header with
|
||||
a bearer token (e.g., a Google ID token).
|
||||
|
||||
These header-generating functions are called just before each request, ensuring
|
||||
that fresh credentials or header values can be used.
|
||||
|
||||
### Configuration
|
||||
|
||||
You can configure these dynamic headers as follows:
|
||||
|
||||
```python
|
||||
from toolbox_llamaindex import ToolboxClient
|
||||
|
||||
async with ToolboxClient(
|
||||
"toolbox-url",
|
||||
client_headers={"header1": header1_getter, "header2": header2_getter, ...}
|
||||
) as client:
|
||||
```
|
||||
|
||||
### Authenticating with Google Cloud Servers
|
||||
|
||||
For Toolbox servers hosted on Google Cloud (e.g., Cloud Run) and requiring
|
||||
`Google ID token` authentication, the helper module
|
||||
[auth_methods](https://github.com/googleapis/mcp-toolbox-sdk-python/blob/main/packages/toolbox-core/src/toolbox_core/auth_methods.py) provides utility functions.
|
||||
|
||||
### Step by Step Guide for Cloud Run
|
||||
|
||||
1. **Configure Permissions**: [Grant](https://cloud.google.com/run/docs/securing/managing-access#service-add-principals) the `roles/run.invoker` IAM role on the Cloud
|
||||
Run service to the principal. This could be your `user account email` or a
|
||||
`service account`.
|
||||
2. **Configure Credentials**
|
||||
- Local Development: Set up
|
||||
[ADC](https://cloud.google.com/docs/authentication/set-up-adc-local-dev-environment).
|
||||
- Google Cloud Environments: When running within Google Cloud (e.g., Compute
|
||||
Engine, GKE, another Cloud Run service, Cloud Functions), ADC is typically
|
||||
configured automatically, using the environment's default service account.
|
||||
3. **Connect to the Toolbox Server**
|
||||
|
||||
```python
|
||||
from toolbox_llamaindex import ToolboxClient
|
||||
from toolbox_core import auth_methods
|
||||
|
||||
auth_token_provider = auth_methods.aget_google_id_token(URL)
|
||||
async with ToolboxClient(
|
||||
URL,
|
||||
client_headers={"Authorization": auth_token_provider},
|
||||
) as client:
|
||||
tools = await client.aload_toolset()
|
||||
|
||||
# Now, you can use the client as usual.
|
||||
```
|
||||
|
||||
## Authenticating Tools
|
||||
|
||||
{{< notice note >}}
|
||||
Always use HTTPS to connect your application with the Toolbox service, especially when using tools with authentication configured. Using HTTP exposes your application to serious security risks.
|
||||
{{< /notice >}}
|
||||
|
||||
Some tools require user authentication to access sensitive data.
|
||||
|
||||
### Supported Authentication Mechanisms
|
||||
Toolbox currently supports authentication using the [OIDC
|
||||
protocol](https://openid.net/specs/openid-connect-core-1_0.html) with [ID
|
||||
tokens](https://openid.net/specs/openid-connect-core-1_0.html#IDToken) (not
|
||||
access tokens) for [Google OAuth
|
||||
2.0](https://cloud.google.com/apigee/docs/api-platform/security/oauth/oauth-home).
|
||||
|
||||
### Configure Tools
|
||||
|
||||
Refer to [these
|
||||
instructions](https://googleapis.github.io/genai-toolbox/resources/tools/#authenticated-parameters) on
|
||||
configuring tools for authenticated parameters.
|
||||
|
||||
### Configure SDK
|
||||
|
||||
You need a method to retrieve an ID token from your authentication service:
|
||||
|
||||
```py
|
||||
async def get_auth_token():
|
||||
# ... Logic to retrieve ID token (e.g., from local storage, OAuth flow)
|
||||
# This example just returns a placeholder. Replace with your actual token retrieval.
|
||||
return "YOUR_ID_TOKEN" # Placeholder
|
||||
```
|
||||
|
||||
#### Add Authentication to a Tool
|
||||
|
||||
```py
|
||||
async with ToolboxClient("http://127.0.0.1:5000") as toolbox:
|
||||
tools = toolbox.load_toolset()
|
||||
|
||||
auth_tool = tools[0].add_auth_token_getter("my_auth", get_auth_token) # Single token
|
||||
|
||||
multi_auth_tool = tools[0].add_auth_token_getters({"auth_1": get_auth_1}, {"auth_2": get_auth_2}) # Multiple tokens
|
||||
|
||||
# OR
|
||||
|
||||
auth_tools = [tool.add_auth_token_getter("my_auth", get_auth_token) for tool in tools]
|
||||
```
|
||||
|
||||
#### Add Authentication While Loading
|
||||
|
||||
```py
|
||||
auth_tool = toolbox.load_tool(auth_token_getters={"my_auth": get_auth_token})
|
||||
|
||||
auth_tools = toolbox.load_toolset(auth_token_getters={"my_auth": get_auth_token})
|
||||
```
|
||||
|
||||
{{< notice note >}}
|
||||
Adding auth tokens during loading only affect the tools loaded within that call.
|
||||
{{< /notice >}}
|
||||
|
||||
### Complete Example
|
||||
|
||||
```py
|
||||
import asyncio
|
||||
from toolbox_llamaindex import ToolboxClient
|
||||
|
||||
async def get_auth_token():
|
||||
# ... Logic to retrieve ID token (e.g., from local storage, OAuth flow)
|
||||
# This example just returns a placeholder. Replace with your actual token retrieval.
|
||||
return "YOUR_ID_TOKEN" # Placeholder
|
||||
|
||||
async with ToolboxClient("http://127.0.0.1:5000") as toolbox:
|
||||
tool = toolbox.load_tool("my-tool")
|
||||
|
||||
auth_tool = tool.add_auth_token_getter("my_auth", get_auth_token)
|
||||
result = auth_tool.call(input="some input")
|
||||
print(result)
|
||||
```
|
||||
|
||||
## Binding Parameter Values
|
||||
|
||||
Predetermine values for tool parameters using the SDK. These values won't be
|
||||
modified by the LLM. This is useful for:
|
||||
|
||||
* **Protecting sensitive information:** API keys, secrets, etc.
|
||||
* **Enforcing consistency:** Ensuring specific values for certain parameters.
|
||||
* **Pre-filling known data:** Providing defaults or context.
|
||||
|
||||
### Binding Parameters to a Tool
|
||||
|
||||
```py
|
||||
async with ToolboxClient("http://127.0.0.1:5000") as toolbox:
|
||||
tools = toolbox.load_toolset()
|
||||
|
||||
bound_tool = tool[0].bind_param("param", "value") # Single param
|
||||
|
||||
multi_bound_tool = tools[0].bind_params({"param1": "value1", "param2": "value2"}) # Multiple params
|
||||
|
||||
# OR
|
||||
|
||||
bound_tools = [tool.bind_param("param", "value") for tool in tools]
|
||||
```
|
||||
|
||||
### Binding Parameters While Loading
|
||||
|
||||
```py
|
||||
bound_tool = toolbox.load_tool("my-tool", bound_params={"param": "value"})
|
||||
|
||||
bound_tools = toolbox.load_toolset(bound_params={"param": "value"})
|
||||
```
|
||||
|
||||
{{< notice note >}}
|
||||
Bound values during loading only affect the tools loaded in that call.
|
||||
{{< /notice >}}
|
||||
|
||||
### Binding Dynamic Values
|
||||
|
||||
Use a function to bind dynamic values:
|
||||
|
||||
```py
|
||||
def get_dynamic_value():
|
||||
# Logic to determine the value
|
||||
return "dynamic_value"
|
||||
|
||||
dynamic_bound_tool = tool.bind_param("param", get_dynamic_value)
|
||||
```
|
||||
|
||||
{{< notice note >}}
|
||||
You don't need to modify tool configurations to bind parameter values.
|
||||
{{< /notice >}}
|
||||
|
||||
## Asynchronous Usage
|
||||
|
||||
For better performance through [cooperative
|
||||
multitasking](https://en.wikipedia.org/wiki/Cooperative_multitasking), you can
|
||||
use the asynchronous interfaces of the `ToolboxClient`.
|
||||
|
||||
{{< notice note >}}
|
||||
Asynchronous interfaces like `aload_tool` and `aload_toolset` require an asynchronous environment. For guidance on running asynchronous Python programs, see [asyncio documentation](https://docs.python.org/3/library/asyncio-runner.html#running-an-asyncio-program).
|
||||
{{< /notice >}}
|
||||
|
||||
```py
|
||||
import asyncio
|
||||
from toolbox_llamaindex import ToolboxClient
|
||||
|
||||
async def main():
|
||||
async with ToolboxClient("http://127.0.0.1:5000") as toolbox:
|
||||
tool = await client.aload_tool("my-tool")
|
||||
tools = await client.aload_toolset()
|
||||
response = await tool.ainvoke()
|
||||
|
||||
if __name__ == "__main__":
|
||||
asyncio.run(main())
|
||||
```
|
||||
Reference in New Issue
Block a user