MetaGPT
Multi-agent framework that simulates a software development team
View on GitHubWhat it solves
Generating complete, working software from a single natural-language spec by distributing the work across specialized agents with defined handoff contracts: not just code generation, but the full engineering process.
Overview
MetaGPT assigns LLM agents to the roles you'd find in a real software team (product manager, system architect, engineer, and QA) and routes a natural-language requirement through their standardized handoffs to produce working code. The workflow is modeled on software engineering process documents: PRD → architecture spec → implementation → test report. ICLR 2025 AFlow work extends it to automated workflow search.
Key facts
- Language
- Python
- License
- MIT
- Maturity
- Community canon
- Maintainer
- FoundationAgents
- Reviewed
- 2026-06-11
Where to start
The Team class is the entry point, it registers roles, routes tasks, and manages the message-passing loop that connects agents across the SDLC.
metagpt/team.pyTechnologies
More in this domain
LangGraph
Stateful multi-agent orchestration for LLM applications
The state management problem in multi-agent systems: how you build agents that pass context across turns without losing the execution thread.
AutoGen
Multi-agent conversation framework from Microsoft Research
How to coordinate multiple specialized agents toward a shared goal without writing low-level orchestration logic by hand.
CrewAI
Role-based multi-agent orchestration for collaborative task execution
The boilerplate problem in multi-agent setup: defining roles, delegation rules, and task routing without writing a custom orchestration graph.