Google’s Agent Development Kit Now Supports Go: Build Scalable AI Agents

Google announced that its Agent Development Kit (ADK) now supports the Go programming language, allowing Go developers to construct and manage AI agents while leveraging Go’s high‑concurrency and strong‑type advantages.

The Go version of ADK is an open‑source toolkit for building modular multi‑agent systems. Agents can be organized hierarchically, and the kit provides debugging, version control, and flexible deployment capabilities.

ADK was designed to help developers who want to tightly integrate advanced AI agents with Google Cloud services.

ADK adopts a “code‑first” development model: all logic, tools, and orchestration are defined in code, which improves flexibility, testability, and version management. It includes a built‑in development UI (ADK web) for testing, evaluating, debugging, and demoing agents.

The ADK web UI is a Node.js + Angular application accessible at localhost:4200. It displays runtime events, trace information, and generated content. Google engineer Daniela Petruzalek explained that the UI is especially useful for teams focused on building agent capabilities, though teams may eventually need custom interfaces that interact with the ADK runtime. She demonstrated a simple front‑end built with HTML/CSS, JavaScript, and a FastAPI (Python) backend.

The Go ADK is part of a broader ADK ecosystem that also includes Java and Python implementations. It bundles OpenAI‑compatible specifications, pre‑built tools, and support for custom functions. Pre‑built tools enable agents to perform tasks such as Gemini web search, code execution, Google Cloud API calls, or integration with third‑party services. Developers can also create custom tools for database queries, document extraction, and more.

ADK supports the A2A protocol for agent‑to‑agent (A2A) interoperability and collaboration. Using A2A, a primary agent can delegate tasks to subordinate agents—whether they run locally or remotely—without exposing internal memory or proprietary logic. The protocol is provided via a separate library that can be extended to support various communication protocols and database back‑ends.

Quick‑start reference: Google ADK sample repository, which contains examples ranging from simple chatbots to complex multi‑agent workflows.

Official documentation: comprehensive guide covering all ADK features and usage.

Conclusion

With Go support, ADK expands its reach to developers who prefer Go’s performance characteristics, offering a full‑stack solution for building, testing, and deploying AI agents that integrate tightly with Google Cloud services.