AutoGen / AG2
Microsoft's conversational multi-agent framework where agents talk to each other in dialogue to solve problems collaboratively.
Last updated: April 26, 2026
Definition
AutoGen, originally from Microsoft Research, models multi-agent systems as conversations: agents (each with their own system prompt and tools) chat with each other to solve problems. The pattern is well-suited to research and creative work where back-and-forth refinement matters. After Microsoft restructured the project in 2024-2025, the active community fork is AG2 (ag2.ai), which preserves and extends the AutoGen architecture. Microsoft's official direction moved to a different framework called AutoGen Studio plus the AgentChat module of Semantic Kernel.
When To Use
AutoGen / AG2 fits research, content, and creative workflows where iterative dialogue between specialized agents adds value. Less suitable for transactional workflows where determinism and audit trails matter more.
Related Terms
Building with AutoGen / AG2?
I've shipped this pattern in real production systems. If you want a second pair of eyes on your architecture, that's what I do.