Skip to content
Agent Development Kit (ADK)
MCP tools
Python
JS
Go
Java
Kotlin
Home
Build Agents
Run Agents
Components
Integrations
Reference
Community
ADK 2.0
Python
JS
Go
Java
Kotlin
Home
Build Agents
Build Agents
Get Started
Get Started
Python
TypeScript
Go
Java
Kotlin
Agents CLI
Installation
Google Cloud
Migrate to ADK
Build your Agent
Build your Agent
Multi-tool agent
Agent team
Code with AI
Agent Config
Agents
Agents
Simple agents
Managed agents
Graph Workflows
Graph Workflows
Graph routes
Data handling
Human input
Dynamic workflows
Multi-Agent Workflows
Multi-Agent Workflows
Collaborative workflows
Template workflows
Template workflows
Sequential workflow
Loop workflow
Parallel workflow
Custom template workflows
Agent routing
Workflow patterns
Models for Agents
Models for Agents
Gemini
Gemma
Claude
Agent Platform hosted
Apigee AI Gateway
Model routing
OpenAI
Ollama
vLLM
LiteLLM
LiteRT-LM
Run Agents
Run Agents
Agent Runtime
Agent Runtime
Web Interface
Web Interface
Visual Builder
Command Line
API Server
Ambient Agents
Resume Agents
Cancel Agent Runs
Runtime Config
Event Loop
Deployment
Deployment
Agent Runtime
Agent Runtime
Standard deployment
agents-cli
Test deployed agents
Cloud Run
GKE
Observability
Observability
Logging
Metrics
Traces
Evaluation
Evaluation
Criteria
User Simulation
Environment Simulation
Custom Metrics
Optimization
Safety and Security
Safety and Security
Components
Components
Technical Overview
Custom Tools
Custom Tools
Function tools
Function tools
Overview
Tool performance
Action confirmations
MCP tools
MCP tools
Table of contents
What is Model Context Protocol (MCP)?
Key considerations
Prerequisites
1. Using MCP servers with ADK agents (ADK as an MCP client) in adk web
McpToolset class
Example 1: File System MCP Server
Step 1: Define your Agent with McpToolset
Step 2: Create an __init__.py file
Step 3: Run adk web and Interact
Example 2: Google Maps Grounding Lite MCP Server
Step 1: Enable the Maps Grounding Lite service on your Google Cloud project
Step 2: Define your Agent with McpToolset for Google Maps Grounding Lite
Step 3: Ensure __init__.py Exists
Step 4: Run adk web and Interact
2. Build an MCP server with ADK tools (MCP server exposing ADK)
Summary of steps
Prerequisites
Step 1: Create the MCP Server Script
Step 2: Implement the Server Logic
Step 3: Test your Custom MCP Server with an ADK Agent
Advanced use cases
Use MCP Tools without adk web
Handling progress updates
Deploy Agents with MCP Tools
Critical Deployment Requirement: Synchronous Agent Definition
Quick Deployment Commands
Agent Runtime
Cloud Run
Deployment Patterns
Pattern 1: Self-Contained Stdio MCP Servers
Pattern 2: Remote MCP Servers (Streamable HTTP)
Pattern 3: Sidecar MCP Servers (GKE)
Connection Management Considerations
Stdio Connections
SSE/HTTP Connections
Production Deployment Checklist
Environment-Specific Configurations
Cloud Run
GKE
Agent Runtime
Troubleshooting Deployment Issues
Further Resources
OpenAPI tools
Authentication
Tool limitations
Artifacts
Artifacts
Skills for Agents
Skills for Agents
App management
App management
Callbacks
Callbacks
Types of callbacks
Callback patterns
Plugins
Agent context
Agent context
Conversational context
Sessions
Sessions
Rewind sessions
Migrate sessions
State
Events
Memory
Context compression
Model context caching
MCP
MCP
A2A Protocol
A2A Protocol
Introduction to A2A
A2A Quickstart (Exposing)
A2A Quickstart (Exposing)
Python
Go
Java
A2A Quickstart (Consuming)
A2A Quickstart (Consuming)
Python
Go
Java
Kotlin
A2A Extension
Live and Voice Agents
Live and Voice Agents
Get started
Get started