At Google Cloud Next 2025, Google unveiled the Agent Development Kit (ADK)—an open-source framework designed to simplify the creation of intelligent, multi-agent systems. ADK empowers developers to build, evaluate, and deploy AI agents capable of reasoning, planning, and collaborating across various tasks and environments.
Key Features of ADK
1. Multi-Agent Architecture
ADK allows for the construction of modular applications by composing multiple specialized agents in a hierarchy, enabling complex coordination and delegation.
2. Flexible Model Integration
Developers can choose from a variety of models, including Google’s Gemini, as well as models accessible via Vertex AI Model Garden. ADK also supports integration with third-party providers like Anthropic, Meta, Mistral AI, and AI21 Labs through LiteLLM.
3. Rich Tool Ecosystem
Agents can be equipped with diverse capabilities using pre-built tools (e.g., Search, Code Execution), Model Context Protocol (MCP) tools, third-party libraries (e.g., LangChain, LlamaIndex), or even other agents as tools (e.g., LangGraph, CrewAI).
4. Integrated Developer Experience
ADK offers a powerful CLI and a visual Web UI for developing, testing, and debugging agents locally. Developers can inspect events, state, and agent execution step-by-step.
5. Built-in Evaluation and Deployment
ADK provides tools to systematically assess agent performance by evaluating both the final response quality and the step-by-step execution trajectory against predefined test cases. Agents can be containerized and deployed anywhere, with optimized integration for Google Cloud’s Vertex AI.
Enhancing Interoperability with Agent2Agent Protocol
Alongside ADK, Google introduced the Agent2Agent (A2A) protocol, an open standard enabling AI agents to communicate and collaborate across different frameworks and vendors. A2A facilitates secure information exchange and coordinated actions among agents, promoting a dynamic, multi-agent ecosystem.
Real-World Applications
Organizations are already leveraging ADK to build sophisticated multi-agent systems. For instance, Revionics is developing agents to assist retailers in setting competitive prices, while Renault Group uses ADK-powered agents to analyze data for strategic EV charger placements.
Getting Started
Developers interested in exploring ADK can access the official documentation and sample agents on the ADK GitHub page. With its comprehensive features and open-source nature, ADK represents a significant step forward in building intelligent, collaborative AI systems.
