Microsoft Kills AutoGen and Replaces It With Agent Framework
Microsoft has moved its 60k-star AutoGen repo into maintenance mode and pushed all new work to the unified Agent Framework 1.0.
Takeaways
- Microsoft has moved AutoGen into maintenance mode: bug fixes only, community-managed from here.
- Microsoft Agent Framework 1.0 is the official successor with stable APIs and long-term support.
- MAF unifies AutoGen and Semantic Kernel into one SDK for Python and .NET, built by the same teams.
- Adds graph-based workflows, checkpointing, human-in-the-loop, OpenTelemetry, and declarative YAML agents.
- Ships with connectors for Foundry, Azure OpenAI, OpenAI, Anthropic, Bedrock, Gemini, and Ollama.
- Alternatives: AG2 community fork for AutoGen continuity, or stay on frozen AutoGen releases.
The most popular multi-agent framework in the Microsoft ecosystem just stopped taking new features. AutoGen, the Python-first agent orchestration library that grew out of Microsoft Research and racked up more than 60,000 GitHub stars, has been officially placed into maintenance mode. The repository will keep receiving bug fixes and security patches, but the roadmap has moved elsewhere.
That elsewhere is Microsoft Agent Framework (MAF), which Microsoft is positioning as the enterprise-ready successor to AutoGen, available as a production release with stable APIs and a commitment to long-term support. If you have production code sitting on top of AutoGen, this is the kind of announcement that reshapes your next quarter of engineering.
What actually changed
The AutoGen README now opens with a maintenance-mode banner. The project will receive no new features or enhancements and is community-managed going forward. Contributions on the main repo are limited to bug fixes, security patches, and documentation improvements. Feature work has moved to the new framework.
Microsoft frames the reason for the consolidation as ending the overlap between two internal efforts. The company was maintaining AutoGen for multi-agent research patterns and Semantic Kernel for enterprise AI integration, with different teams, different APIs, and increasing feature duplication. MAF collapses those into one SDK.
What Microsoft Agent Framework actually is
MAF is neither a rebrand nor a wrapper. The same teams that built AutoGen and Semantic Kernel built the successor. It ships as