Blueprint · intermediate · 9 steps
Build an Agent with Google ADK
Build a research agent with Google's Agent Development Kit and Gemini — custom tools, sessions, and one-command deploy to Cloud Run.
All steps
01Step 1: What We're BuildingA research agent powered by Google's Agent Development Kit (ADK) and Gemini that can search the web, perform calculations, save notes, and deploy to Cloud Run with a single command.→02Step 2: Project SetupCreate the Python project, install Google ADK, and configure your Gemini API key.→03Step 3: Your First AgentDefine a minimal ADK agent with Gemini, run it with the built-in CLI and web UI, and understand the core Agent class.→04Step 4: Add Custom ToolsDefine custom Python functions as tools — a calculator and a note saver — and connect them to your agent so Gemini can call them.→05Step 5: The Agent LoopUnderstand the Runner, Events, and the execution loop that powers your agent — then build a Python script that runs multi-turn conversations.→06Step 6: Web SearchAdd Google Search to your agent using ADK's built-in tool — no extra API keys, no setup, just one import.→07Step 7: Sessions and MemoryAdd session management and state persistence so your agent remembers context across conversations — using ADK's built-in session services.→08Step 8: Deploy to Cloud RunDeploy your research agent to Google Cloud Run with a single ADK command — get a live HTTPS endpoint anyone can access.→09Step 9: What's NextProduction hardening, multi-agent systems, and where to go from here with your ADK agent.→