Technology RadarTechnology Radar

AutoGen / AG2

agentsopen-source
Assess

Microsoft's open-source multi-agent framework and its community fork — significant research influence but a fragmented story.

Why It Matters

AutoGen (originally Microsoft Research) pioneered conversational multi-agent patterns where agents collaborate through structured conversations. AG2 is the community-driven fork that emerged when Microsoft refactored AutoGen into a new architecture (AutoGen 0.4+). Together, they represent one of the most influential approaches to multi-agent orchestration, with an agent gallery and extension ecosystem. The framework includes an AgentChat layer and a growing collection of pre-built agent patterns.

Strengths

  • Strong research foundation — Microsoft Research's multi-agent patterns are well-published and cited
  • Agent gallery provides pre-built, composable agent patterns
  • Extension system allows community-contributed tools and capabilities
  • AutoGen Studio provides a visual builder for non-developers
  • Active academic and research community driving innovation

Limitations

  • The AutoGen/AG2 split has fragmented the community and created confusion
  • AutoGen 0.4's architectural rewrite broke backward compatibility with 0.2
  • No managed hosting or deployment — you run everything yourself
  • Agent gallery is research-quality, not production-quality

Risks

  • The AutoGen/AG2 fork drama poisoned community trust; developers don't know which version to invest in
  • Microsoft's commitment to AutoGen is uncertain — it's a research project, not a product with a business model
  • The 0.2 → 0.4 rewrite was effectively a new framework, invalidating existing tutorials, integrations, and agent definitions
  • AutoGen Studio's visual builder is impressive for demos but produces agents that are hard to customize or debug
  • Multi-agent conversations are non-deterministic by design, making AutoGen agents difficult to test and verify in production