Google Unveils Open-Source Framework for Building Reliable AI Agents
Google has released its Agent Development Kit (ADK), an open-source multi-agent framework designed to build reliable and production-ready artificial intelligence (AI) agent systems. The toolkit is optimized for Google's Gemini models but can work with other language models via LiteLLM.
The ADK shifts from a rigid, top-down hierarchy to a powerful graph-based execution engine that mixes deterministic code with adaptive AI. It officially supports Python and Go for its 2.0 graph architecture, alongside TypeScript for standard implementation.
ADK 2.0 replaces the hierarchical model of ADK 1.x with a directed execution graph and provides dual orchestration: graph-based workflows with explicitly mapped-out node/edge routing and dynamic workflows written in plain code. The toolkit also bundles a formal 'Task API' for structured agent-to-agent delegation.
The system includes built-in human-in-the-loop (HITL) functionality, which treats human gates as native primitives within the graph runtime. This allows the system to halt execution, await human feedback or approval, and resume from any specified checkpoint.