GPTeam is an open-source multi-agent simulation in which a team of GPT-4 powered agents collaborates to reach shared goals. Each agent has its own role, memory and view of a shared workspace of tasks and messages; the agents plan, divide work, comment on each other’s progress and revise — a small virtual company that runs itself in front of you. The project has around 1,700 stars and is actively maintained, with code last updated in 2026.
GPTeam was built to explore how far LLM agents get when they cooperate rather than work alone. You define a mission and a roster of agents, and the simulation runs in a shared environment where agents post updates, claim tasks and build on each other’s output. A hosted web app mirrors the open-source version, and the architecture was featured in LangChain’s blog as a reference multi-agent design.
Free. GPTeam is open source under the MIT license; you self-host it and pay only for your own OpenAI API usage, and the hosted demo is free to try.
Agent researchers and developers studying multi-agent collaboration, educators demonstrating autonomous-agent concepts, and tinkerers who want a readable reference implementation of an agent team.