ChatDev
Simulate a full software company where every role (CEO to QA) is an LLM agent
View on GitHubWhat it solves
The quality gap in single-agent code generation: having specialized reviewer, architect, and tester agents catch issues that a lone code-generating agent would consistently miss.
Overview
ChatDev from OpenBMB models a software development company as a multi-agent simulation. A CEO agent, CTO agent, programmer agents, and a QA agent each have defined responsibilities and communicate through a 'chat chain' of structured dialogues. Hand it a product idea in natural language; the simulated company designs, builds, and tests a working prototype. The ExperiencePool extension lets agents accumulate and share institutional knowledge across runs.
Key facts
- Language
- Python
- License
- Apache 2.0
- Maturity
- Stable
- Maintainer
- OpenBMB
- Reviewed
- 2026-06-11
Where to start
Running the default task is the fastest way to see the chat chain in action, watch the role-switching logs to understand the communication protocol before reading the framework code.
run.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.