This documentation provides a complete end-to-end guide for building advanced agents using the Google AI Developer Kit (ADK). It explains Custom Tools, Long-Running Tools, MCP Servers, Function Tools, OpenAPI Tools, and Built-in Tools such as Google Search and the Gemini Code Executor.
You’ll learn how agents use Python code execution for reliable math, how to integrate third-party systems through the Model Context Protocol (MCP), how to build Sub-Agents and Agent Tools, and how to connect to external MCP servers using stdio. The guide includes a full walkthrough of the Shipping Coordinator workflow using ToolContext, confirmation events, resumability, invocation IDs, and long-running tool design.
You also get complete annotated code for approval-based workflows, event parsing, session handling, and step-by-step explanation of multi-tool coordination. This is the ultimate reference for developers building enterprise-grade, reliable, and secure ADK agents.
