AutoGen
Microsoft's open-source framework for multi-agent conversation — agents talk to each other to solve problems collaboratively, with explicit support for code execution and human-in-the-loop.
AutoGen was Microsoft Research's contribution to the multi-agent space. The core pattern: define agents (each with its own system prompt and tools), let them converse, intervene when needed. The "conversation" metaphor is the differentiator — agents literally talk to each other in natural language until they reach a conclusion.
Strengths: explicit human-in-the-loop support, strong code execution, research-grade flexibility. Used heavily in academic AI research. Weaknesses: conversation can be inefficient; agents can over-talk; less production polish than LangGraph or CrewAI.
In 2026, AutoGen is most popular in research contexts and at Microsoft-aligned enterprises. For most production work, LangGraph or vendor SDKs are the better choice. AutoGen-Studio (a GUI on top) makes it more accessible.
Frequently asked
AutoGen vs CrewAI?+
AutoGen for research and conversational multi-agent. CrewAI for role-based crews. Both are open source, both work — pick by which mental model fits your problem.
Is AutoGen still maintained?+
Yes, by Microsoft Research. AutoGen 0.4 was a major rewrite. Production users should expect occasional API changes; pin versions for stability.