ADK Package

Overview

The tbadk package provides a Go interface to the MCP Toolbox service, enabling you to load and invoke tools from your own applications.

Installation

go get github.com/googleapis/mcp-toolbox-sdk-go/tbadk

This SDK is supported on Go version 1.24.4 and higher.

Quickstart

Here’s a minimal example to get you started. Ensure your Toolbox service is running and accessible.

package main

import (
  "context"
  "fmt"
  "github.com/googleapis/mcp-toolbox-sdk-go/tbadk"
)

func quickstart() string {
  inputs := map[string]any{"location": "London"}
  client, err := tbadk.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)
  }
  // pass the tool.Context as ctx into the Run() method
  result, err := tool.Run(ctx, inputs)
  if err != nil {
    return fmt.Sprintln("Could not invoke tool", err)
  }
  return fmt.Sprintln(result["output"])
}

func main() {
  fmt.Println(quickstart())
}

Usage

Import and initialize a Toolbox client, pointing it to the URL of your running Toolbox service.

import "github.com/googleapis/mcp-toolbox-sdk-go/tbadk"

client, err := tbadk.NewToolboxClient("http://localhost:5000")

All interactions for loading and invoking tools happen through this client.

Transport Protocols

The SDK supports multiple transport protocols. 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.

Supported Protocols

Example

import (
    "github.com/googleapis/mcp-toolbox-sdk-go/core"
    "github.com/googleapis/mcp-toolbox-sdk-go/tbadk"
)

// Initialize with the native Toolbox protocol
client, err := tbadk.NewToolboxClient(
    "http://localhost:5000",
    core.WithProtocol(core.Toolbox),
)

// Initialize with the MCP Protocol 2025-03-26
client, err := tbadk.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:

// 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)

LoadToolset returns a slice of the ToolboxTool structs ([]ToolboxTool).

Load a single tool

Loads a specific tool by its unique name. This provides fine-grained control.

tool, err = client.LoadTool("my-tool", ctx)

Invoking Tools

Once loaded, tools behave like Go structs. You invoke them using Run method by passing arguments corresponding to the parameters defined in the tool’s configuration within the Toolbox service.

tool, err = client.LoadTool("my-tool", ctx)
inputs := map[string]any{"location": "London"}
// Pass the tool.Context as ctx to the Run() function
result, err := tool.Run(ctx, inputs)

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, 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:

import (
  "github.com/googleapis/mcp-toolbox-sdk-go/core"
  "github.com/googleapis/mcp-toolbox-sdk-go/tbadk"
)

tokenProvider := func() string {
  return "header3_value"
}

staticTokenSource := oauth2.StaticTokenSource(&oauth2.Token{AccessToken: "header2_value"})
dynamicTokenSource := core.NewCustomTokenSource(tokenProvider)

client, err := tbadk.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 provides utility functions.

Step by Step Guide for Cloud Run

1. Configure Permissions: Grant the roles/run.Runr 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.
   • 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

   import (
     "github.com/googleapis/mcp-toolbox-sdk-go/core"
     "github.com/googleapis/mcp-toolbox-sdk-go/tbadk"
     "context"
   )
   
   ctx := context.Background()
   
   token, err := core.GetGoogleIDToken(ctx, URL)
   
   client, err := tbadk.NewToolboxClient(
     URL,
     core.WithClientHeaderString("Authorization", token),
   )
   
   // Now, you can use the client as usual.
   

Authenticating Tools

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.

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 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).

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
}

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.

ctx := context.Background()

client, err := tbadk.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.

ctx := context.Background()

client, err := tbadk.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.

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"),
)

Complete Authentication Example

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 := tbadk.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.Run(ctx, inputs)

  fmt.Println(result)
}

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.

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.

ctx := context.Background()

client, err := tbadk.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.

client, err := tbadk.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.

boundTool, err := client.LoadTool("my-tool", ctx, core.WithBindParamString("param", "value"))

// OR

boundTool, err := client.LoadToolset("", ctx, core.WithBindParamString("param", "value"))

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.

getDynamicValue := func() (string, error) { return "req-123", nil }

dynamicBoundTool, err := tool.ToolFrom(core.WithBindParamStringFunc("param", getDynamicValue))

Default Parameters

Tools defined in the MCP Toolbox server can specify default values for their optional parameters. When invoking a tool using the SDK, if an input for a parameter with a default value is not provided, the SDK will automatically populate the request with the default value.

tool, err = client.LoadTool("my-tool", ctx)

// If 'my-tool' has a parameter 'param2' with a default value of "default-value",
// we can omit 'param2' from the inputs.
inputs := map[string]any{"param1": "value"}

// The invocation will automatically use param2="default-value" if not provided
result, err := tool.Run(ctx, inputs)

Using with ADK Go

After altering the tool to your needs, type-assert the ToolboxTool and pass it to the LLM agent.

For a single tool

toolboxtool, err = client.LoadTool("my-tool", ctx)

// <Bind parameters & add authentication here>

llmagent, err := llmagent.New(llmagent.Config{
  Name:        "assistant",
  Model:       model,
  Description: "Agent to answer questions.",
  Tools:       []tool.Tool{&toolboxtool},
})

For a toolset

toolboxtools, err := client.LoadToolset("", ctx)

// <Bind parameters & add authentication here>

toolsList := make([]tool.Tool, len(toolboxtools))
	for i := range toolboxtools {
		toolsList[i] = &toolboxtools[i]
	}

llmagent, err := llmagent.New(llmagent.Config{
  Name:        "assistant",
  Model:       model,
  Description: "Agent to answer questions.",
  Tools:       toolsList,
})

The reason we have to type assert it before passing it to ADK Go, is because it requires a generic tool.Tool interface. You can always convert it back to ToolboxTool format to access the specialized methods.

Using with Orchestration Frameworks

To see how the MCP Toolbox Go SDK works with orchestration frameworks, check out these end-to-end examples given below.

//This sample contains a complete example on how to integrate MCP Toolbox Go SDK with ADK Go using the tbadk package.
package main

import (
	"context"
	"fmt"
	"log"
	"os"

	"github.com/googleapis/mcp-toolbox-sdk-go/tbadk"
	"google.golang.org/adk/agent"
	"google.golang.org/adk/agent/llmagent"
	"google.golang.org/adk/model/gemini"
	"google.golang.org/adk/runner"
	"google.golang.org/adk/session"
	"google.golang.org/adk/tool"
	"google.golang.org/genai"
)

func main() {
	genaiKey := os.Getenv("GEMINI_API_KEY")
	toolboxURL := "http://localhost:5000"
	ctx := context.Background()

	// Initialize MCP Toolbox client
	toolboxClient, err := tbadk.NewToolboxClient(toolboxURL)
	if err != nil {
		log.Fatalf("Failed to create MCP Toolbox client: %v", err)
	}

	toolsetName := "my-toolset"
	toolset, err := toolboxClient.LoadToolset(toolsetName, ctx)
	if err != nil {
		log.Fatalf("Failed to load MCP toolset '%s': %v\nMake sure your Toolbox server is running.", toolsetName, err)
	}

	// Create Gemini model
	model, err := gemini.NewModel(ctx, "gemini-2.5-flash", &genai.ClientConfig{
		APIKey: genaiKey,
	})
	if err != nil {
		log.Fatalf("Failed to create model: %v", err)
	}

	tools := make([]tool.Tool, len(toolset))
	for i := range toolset {
		tools[i] = &toolset[i]
	}

	llmagent, err := llmagent.New(llmagent.Config{
		Name:        "hotel_assistant",
		Model:       model,
		Description: "Agent to answer questions about hotels.",
		Tools:       tools,
	})
	if err != nil {
		log.Fatalf("Failed to create agent: %v", err)
	}

	appName := "hotel_assistant"
	userID := "user-123"

	sessionService := session.InMemoryService()
	resp, err := sessionService.Create(ctx, &session.CreateRequest{
		AppName: appName,
		UserID:  userID,
	})
	if err != nil {
		log.Fatalf("Failed to create the session service: %v", err)
	}
	session := resp.Session

	r, err := runner.New(runner.Config{
		AppName:        appName,
		Agent:          llmagent,
		SessionService: sessionService,
	})
	if err != nil {
		log.Fatalf("Failed to create runner: %v", err)
	}

	query := "Find hotels with Basel in its name."

	fmt.Println(query)
	userMsg := genai.NewContentFromText(query, genai.RoleUser)

	streamingMode := agent.StreamingModeSSE
	for event, err := range r.Run(ctx, userID, session.ID(), userMsg, agent.RunConfig{
		StreamingMode: streamingMode,
	}) {
		if err != nil {
			fmt.Printf("\nAGENT_ERROR: %v\n", err)
		} else {
			if event.LLMResponse.Content != nil {
				for _, p := range event.LLMResponse.Content.Parts {
					if streamingMode != agent.StreamingModeSSE || event.LLMResponse.Partial {
						fmt.Print(p.Text)
					}
				}
			}
		}
	}
	fmt.Println()
}