AutoGen

Microsoft's multi-agent framework appears in three source notes as a member of a recurring list rather than as a subject in its own right: LangGraph, CrewAI, AutoGen, sometimes joined by Semantic Kernel, OpenAI Agents SDK, AWS Bedrock and Google ADK. That enumeration is the point. In each case the writer names the framework cluster in order to argue that something else matters more.

Tony Seale, writing on 2026-04-17, sets the orchestration frameworks against the pattern he sees among advanced practitioners: a capable model in a capable harness, as with «Claude Code», Codex, OpenClaw and Hermes. He quotes Anthropic to the effect that every harness component encodes an assumption about what the model cannot do alone, and that such assumptions age fast. Scaffolding built for a limited model handicaps a stronger one. Seale calls frameworks a transitory phase; what an organization owns is its ontology.

Antoine Habert reaches a related conclusion from two years of production deployment in a European fintech, covering 2023 to 2025. His account of 2024 is that framework proliferation, «AutoGen» among them, left the harder dimensions underdeveloped: reasoning transparency, structured organizational memory, genuine cognitive cooperation between agents, and scalable supervision. The Wenvision platform note repeats the gap in plainer terms: building a basic agent with LangChain or Autogen is comparatively easy, and enterprise deployment demands governance, auditability, cost control and compliance instead.

The attributes recorded here call «AutoGen» a framework for prototyping AI agents. None of the three notes disputes that; they dispute what prototyping is worth.