AutoGen vs MetaGPT
Multi-agent conversation framework for building LLM applications with cooperating agents. — versus — The original "AI software company" multi-agent framework — role-assigned agents (PM, architect, engineer) turn a one-line requirement into PRD, design, and code.
Same job — multi-agent LLM apps — opposite philosophy: AutoGen is a conversation substrate you shape freely; MetaGPT hard-codes the workflow as company SOPs.
| AutoGen | MetaGPT | |
|---|---|---|
| Stars | 60k | 69k |
| Forks | 9.0k | 8.8k |
| Language | Python | Python |
| License | CC-BY-4.0 | MIT |
| Last activity | 3 months ago | 5 months ago |
| Topics | agents, evals | agents, orchestration |
| Curated connections | 9 | 3 |
AutoGen — the curator's take
Multi-agent conversation framework for building LLM applications with cooperating agents.
MetaGPT — the curator's take
MetaGPT is the canonical demonstration that multi-agent choreography works: encode a software company's SOPs as roles and watch one prompt become a PRD, a design doc, and running code. Study it for that — the role/SOP pattern shows up everywhere now. But be honest about 2026: agentic coding tools (Claude Code, Codex) have eaten the "write my app from a prompt" job, and the repo's cadence has slowed noticeably (last push Jan 2026) while the team focuses on its commercial MGX product. For production pipelines you control, reach for LangGraph; for lightweight role crews, CrewAI. Reach for MetaGPT to learn from the most complete SOP-driven design in the wild.