Google's Agent Development Kit (ADK) Expands Language Support to Five Languages

Expanded Language Support for Agent Development

Google has announced the release of its Agent Development Kit (ADK), which now supports five programming languages: Python, Go, TypeScript, Java, and Kotlin. This update addresses the common limitation in agent frameworks that often assume a Python-only development stack. The ADK allows developers to build agents in their chosen language, removing the need to create separate Python services or custom orchestration logic for non-Python environments.

Unified Architecture and Interoperability

Each language-specific version of the ADK is designed for its respective runtime and type system, while maintaining a unified architecture. Interoperability is achieved through open protocols. The ADK includes a Model Context Protocol (MCP) client in every language, enabling custom tools exposed as MCP servers to be called from any supported language. For multi-agent systems, the ADK supports the Agent2Agent (A2A) protocol, which facilitates communication between remote agents over HTTP, even if they are written in different languages or frameworks.

Getting Started and Key Features

Developers can begin by running a one-time setup command to configure Google Cloud authentication and install the CLI. The ADK also allows setting up first-party ADK skills in coding agents like Antigravity, Claude Code, or Codex. Both the ADK for Python and ADK for Kotlin repositories include agent builder skills in their source trees. Additionally, the adk.dev/llms.txt resource provides on-demand API documentation across all five languages for any coding agent. Coding-agent skills scaffold projects, while MCP and A2A connect the different language runtimes.

Python as Reference Implementation

The ADK for Python serves as the reference implementation. It offers the broadest set of Google Cloud integrations, including BigQuery, Spanner, and Bigtable, along with model adapters for Gemini, Anthropic, and OpenAI. In ADK for Python, developers can pass a plain Python function into the tools parameter, and ADK automatically inspects its signature, type hints, and docstring at startup to construct the tool declaration and Pydantic parameter schema.