ag2

Why we included this project

AG2, the framework formerly known as AutoGen, is a Python toolkit for building applications where several AI agents work together instead of answering one-off prompts. Agents can call tools, pause for human input mid-run, and coordinate through multi-agent conversation patterns, and the async-first API slots neatly into modern Python codebases. The 1.0 release rebuilt the core around a protocol-driven design while keeping the classic AutoGen-style classes maintained in a separate package, so teams can pick whichever approach fits their workflow. With provider-agnostic configuration and a large collection of example applications, there is plenty to draw on once you start wiring real workloads together.