Google Agent Development Kit

Google's Agent Development Kit takes a code-first, model-agnostic approach to building, evaluating, and deploying AI agents, and it is the same framework that powers Agent Engine on Google Cloud. Python is the primary implementation, installed with pip install google-adk on Python 3.10 or newer, with official TypeScript, Go, Java, and Kotlin versions maintained in sibling repositories. Version 2.0 introduced graph-based workflows with routing, fan-out and fan-in, loops, retries, state management, and human-in-the-loop steps, alongside a Task API for structured agent-to-agent delegation, built-in evaluation tooling, and a local web UI for testing. Tools plug in as custom functions, OpenAPI specs, or MCP servers, and the A2A protocol is supported for exposing and consuming remote agents. Gemini gets native integration while adapters cover Claude, Gemma, and locally served models through Ollama or vLLM. Deployment targets include Agent Runtime, Cloud Run, and GKE, or any container platform. The project is Apache-2.0, ships updates roughly biweekly, holds about 20.9k stars, and documents everything at adk.dev.