ADK Go is the Go edition of Google’s Agent Development Kit: an open-source, code-first toolkit for building, evaluating and deploying sophisticated AI agents in the Go programming language. It brings the same agent framework Google offers for Python and Java to Go developers, with about 8,900 stars on GitHub and active development in 2026.
ADK Go lets you define agents in code with full control over their behavior, tools and orchestration. It supports multi-agent systems where specialized agents collaborate and hand off work, integrates with the Model Context Protocol (MCP) for tool access, and speaks the Agent2Agent (A2A) protocol for interoperability with other agent systems. While it is optimized for Google’s Gemini models and Vertex AI, the framework is model-agnostic in design. Built-in evaluation tooling helps you test agent quality before deployment, and agents can be deployed to environments such as Google Cloud Run or Vertex AI Agent Engine. The toolkit covers the full agent lifecycle rather than just the runtime: sessions and memory give agents short- and long-term context, callbacks and plugins let you hook into each step of an agent’s reasoning, and streaming support enables responsive, token-by-token user experiences. Teams already invested in Go services can therefore add agentic features inside their existing stack and deployment practices, instead of introducing a separate Python platform just for AI features.
ADK Go is free and open source under the Apache-2.0 license. You only pay for the models and cloud infrastructure your agents use.
ADK Go is for Go developers and Google Cloud teams who want a first-class, code-driven framework for building and shipping multi-agent systems.