Product / UX / Orchestration & agents

AutoGen

Microsoft's framework for agentic AI built around multi-agent conversation, where model-backed agents collaborate, critique each other and run tools and code. Its event-driven redesign and code-execution focus distinguish it from graph- or role-based frameworks. Microsoft now directs new work to the Microsoft Agent Framework, which merges AutoGen with Semantic Kernel.

The repository carries separate license files for code and documentation, which makes GitHub's license field misleading; the openness axis reads the code file. Verified 2026-08-13 via the microsoft/autogen repository and its LICENSE-CODE.

Openness

5 high confidence

5.0

license: MIT(code,via LICENSE-CODE)
docs: CC-BY-4.0
source: public
core-gated: ungated

Code is MIT and fully open; the GitHub-surfaced CC-BY-4.0 is only the docs license. The whole framework is published with no enterprise directory and no paid tier.

https://github.com/microsoft/autogen/blob/main/LICENSE-CODE recorded 2026-08-13
MIT License text, "Copyright (c) Microsoft Corporation"
https://github.com/microsoft/autogen recorded 2026-08-13
README states the split - documentation under LICENSE, "a license to any code in the repository under the MIT License, see the LICENSE-CODE file"; full framework source published, 60.4k stars, 9.1k forks, no enterprise directory or paid tier

Adoption

4 medium confidence

4.0

1,058,837 PyPI downloads in the trailing 30 days for autogen-agentchat. Bands at level 4 (1M-10M) on the software and model adoption scale. Trajectory is flat to declining as the project enters maintenance.

https://pypistats.org/api/packages/autogen-agentchat/recent recorded 2026-08-12
1,058,837 downloads in the trailing 30 days for autogen-agentchat

Capability

4 medium confidence

4.0

Strong multi-agent conversation model; less durable-execution depth than LangGraph and now in maintenance.

https://github.com/microsoft/autogen recorded 2026-08-13
multi-agent conversation orchestration, team patterns, tool use and code execution published in full; no benchmark figure

Verified 2026-08-12