AutoGen is a Python-based framework developed by Microsoft for building multi-agent AI applications that operate autonomously or in collaboration with humans. The framework is currently in maintenance mode and no longer receives new features or enhancements, with Microsoft recommending that new projects use the Microsoft Agent Framework instead. Existing AutoGen users are encouraged to migrate using the provided migration guide, though the framework remains available for continued use by current projects.
The framework employs a layered and extensible architecture with clearly divided responsibilities across different abstraction levels. The Core API implements message passing, event-driven agents, and both local and distributed runtime capabilities, with support for cross-language functionality across Python and .NET. The AgentChat API provides a simpler, more opinionated interface for rapid prototyping and is most familiar to users of AutoGen v0.2, supporting common multi-agent patterns including two-agent chat and group chat scenarios. The Extensions API enables both first-party and third-party extensions to continuously expand framework capabilities, including specific implementations of LLM clients such as OpenAI and AzureOpenAI, along with capabilities like code execution.
The ecosystem includes two essential developer tools. AutoGen Studio provides a no-code graphical interface for building multi-agent applications without requiring code writing, though it is explicitly not intended for production use and is designed for rapid prototyping and demonstration purposes. AutoGen Bench offers a benchmarking suite for evaluating agent performance. The framework has also enabled the creation of specialized applications such as Magentic-One, a state-of-the-art multi-agent team built using the AgentChat and Extensions APIs that handles tasks requiring web browsing, code execution, and file handling.
According to GitGenius activity tracking, the repository shows significant community engagement with 2515 tracked issues and pull requests. The median response latency across these items is 0.0 hours, though the mean response latency is 1056.4 hours, indicating variable response times. The most active issue labels are needs-triage with 741 occurrences, version 0.2 with 553 occurrences, and proj-studio with 230 occurrences. The most active contributors tracked by GitGenius are ekzhu with 2700 events, rysweet with 2267 events, and victordibia with 1168 events. The repository shares overlapping contributors with microsoft/vscode, microsoft/typescript, and rust-lang/rust, indicating cross-project collaboration within the Microsoft ecosystem.
The framework requires Python 3.10 or later for installation. AutoGen originated from Microsoft Research and pioneered experimental multi-agent orchestration patterns that influenced the broader community. While the framework is no longer receiving active development, community support remains available through a Discord server and GitHub Discussions, with the understanding that responses may be limited given the community-managed status.
