An open-source, code-first Python framework for building, evaluating, and deploying sophisticated AI agents with flexibility and control.
The Agent Development Kit (ADK) 2.0 is an open-source, code-first Python framework for building, evaluating, and deploying sophisticated AI agents. Version 2.0 introduces a Workflow Runtime — a graph-based execution engine for composing deterministic execution flows for agentic apps, with support for routing, fan-out/fan-in, loops, retry, state management, dynamic nodes, human-in-the-loop, and nested workflows. It also adds a Task API for structured agent-to-agent delegation with multi-turn task mode, single-turn controlled output, mixed delegation patterns, human-in-the-loop, and task agents as workflow nodes. Agents and workflows can be run locally via an interactive CLI or a web UI.
Building, evaluating, and deploying sophisticated AI agents using a code-first Python framework, including composing agentic workflows and agent-to-agent task delegation.
ADK is a code-first framework for building and deploying sophisticated AI agents. While the entrypoint code shows no direct autonomy signals (no shell/subprocess/exec in the snippet), the README explicitly describes a graph-based Workflow Runtime supporting routing, fan-out/fan-in, loops, retry, dynamic nodes, and nested workflows, plus a Task API for agent-to-agent delegation. These are agentic multi-step execution patterns that run without per-step human confirmation by default. Human-in-the-loop is offered as an optional feature within workflows/tasks, not a default gate, indicating autonomous operation. The surface analysis also notes the framework executes arbitrary code. This places the framework at Tier 4 — multi-step autonomous action — rather than a purely advisory tier.