GooseTeam

What is GooseTeam?

GooseTeam is an MCP server that enables Goose agents to collaborate by providing tools for agent registration, task management, and message exchange. It is designed for users of the Goose CLI who want multiple agents to work together on projects.

How to use GooseTeam?

Clone the repository, install dependencies with npm install, then build with npm run build. Start the MCP proxy with npm run mcp-proxy and configure Goose to connect via goose configure with a remote extension pointing to http://localhost:3001/sse. Launch agents using npm run agent (text-based protocol) or npm run agent:chart (Mermaid chart protocol).

Key features of GooseTeam

• Agent registration with unique ID and color.
• Message storage and retrieval.
• Task creation, assignment, and completion.
• Agent wait mechanism for loop control.
• MCP proxy for multi‑client connectivity.

Use cases of GooseTeam

• Coordinate a team of agents to complete complex projects.
• Test an LLM’s ability to follow a procedural loop protocol.
• Simulate multi‑agent workflows with a Project Coordinator role.
• Evaluate different models for tool‑use reliability.

FAQ from GooseTeam

Which LLMs work well with GooseTeam?

Currently only google/gemini-2.0-flash-001 (via OpenRouter) and openai/gpt-4o reliably stay in the loop. Other models often end the message early.

What are the runtime dependencies?

You must have the Goose CLI installed and a Node.js environment to run the npm scripts.

How does the MCP Proxy work?

The STDIO-based server is wrapped by an SSE proxy on port 3001, allowing multiple clients to connect to the same server instance.

What happens if a model fails to follow the protocol?

The agent may exit with a “message loop ended” error. You can test with npm run agent:test to verify the model’s suitability.