Repolode
Submit
agentscope-ai

agentscope-ai/AgentTeams

An open-source multi-agent runtime that coordinates Manager and Worker AI agents in Matrix chat rooms, with humans able to watch and intervene.

View on GitHubWebsite
Stars
5,845
Forks
714
Open issues
248
Last commit
2 days ago

What is it?

AgentTeams is a collaborative multi-agent runtime built on a Manager-Workers architecture. A Manager agent creates and assigns tasks to Worker agents, and all communication happens in Matrix rooms where the user, the Manager and the relevant Workers are present, so every exchange is visible and can be interrupted.

The project orchestrates agent containers rather than implementing agent logic itself. It supports OpenClaw, QwenPaw and Hermes Workers in the same room, along with an experimental DeepSeek Harness Worker. A bundled Higress AI gateway holds real API keys so Workers only see consumer tokens, and a MinIO shared file system is used for exchanging information between agents.

Installation runs through a single shell or PowerShell script with Docker, or through a Helm chart on Kubernetes that bundles the gateway, Tuwunel Matrix server, MinIO and an AgentTeams controller configured via CRDs.

Key features

  • 01Manager agent creates Workers and assigns tasks through chat.
  • 02Multiple runtimes (OpenClaw, QwenPaw, Hermes) can work in the same Matrix room.
  • 03Real API keys stay in the Higress gateway, so Workers only hold consumer tokens.
  • 04Built-in Matrix server and Element Web client require no bot application setup.
  • 05Kubernetes deployment via Helm with CRD-based Worker, Team and Human resources.

Use cases

  • Running a team of AI agents that split up software tasks while a human watches every message.
  • Coordinating agents that need to share files without repeatedly passing large context in prompts.
  • Hosting a private multi-agent setup on your own machine or cluster where API keys are not exposed to agents.

Topics

#multi-agent#ai-agent#matrix#kubernetes#llm-gateway#human-in-the-loop#orchestration#agent-teams#openclaw

FAQ

What is AgentTeams?
An open-source multi-agent runtime that coordinates Manager and Worker AI agents in Matrix chat rooms, with humans able to watch and intervene. AgentTeams is a collaborative multi-agent runtime built on a Manager-Workers architecture. A Manager agent creates and assigns tasks to Worker agents, and all communication happens in Matrix rooms where the user, the Manager and the relevant Workers are present, so every exchange is visible and can be interrupted.
Is AgentTeams free to use? What license does it have?
AgentTeams is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is AgentTeams still maintained?
Actively maintained. The last commit to AgentTeams was on Oct 20, 2026.
What can I use AgentTeams for?
Running a team of AI agents that split up software tasks while a human watches every message. Coordinating agents that need to share files without repeatedly passing large context in prompts. Hosting a private multi-agent setup on your own machine or cluster where API keys are not exposed to agents.
Where can I get AgentTeams?
The source code and installation instructions are on GitHub: https://github.com/agentscope-ai/AgentTeams

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects