MCP HubMCP Hub
cliffhall

GooseTeam

by: cliffhall

Look, a flock of geese! An MCP server and protocol for Goose agent collaboration.

37created 15/02/2025
Visit
Goose
protocol

đź“ŚOverview

Purpose: To enable collaboration among AI agents through a structured protocol and extension that facilitates task management and messaging.

Overview: GooseTeam provides a robust framework that allows AI agents to work together efficiently by following defined protocols. It includes both natural language instructions and a mermaid chart for streamlined interaction. The system features an MCP (Model Context Protocol) server extension, which supports agent registration, message handling, task management, and allows for seamless collaboration among multiple agents.

Key Features:

  • Protocol Formats: Offers two formats—natural language for human readability and mermaid markdown for structured flowcharts, enabling flexibility in conveying agent instructions.

  • MCP Server Extension: Facilitates agent registration, task assignment, and message management, allowing agents to communicate effectively and manage collaborative projects.

  • Agent Operations: Includes functionalities like waiting for specified periods before acting and managing tasks dynamically through a project coordinator, enhancing operational efficiency.

GooseTeam

Enabling Goose Agents to Collaborate

Protocol

An operational procedure for agents to follow, expressed in two flavors:

  • Natural Language: Streamlined, human-readable instructions about how to behave and use the tools.
  • Mermaid Markdown: A concise flowchart for expected behavior, described in an efficient text format.

Extension

Goose extension (MCP Server) that provides tools for the agents to use for tasks and messaging.

  • Agent Registration: Facilitates agent registration with unique ID assignments and message sending.
  • Message Management: Stores messages from agents, making them available for retrieval.
  • Task Management: Project Coordinator role creates and assigns tasks and roles to other agents.
  • Agent Waiting: Allows connected agents to wait for a specified period before taking another action.
  • Remote Server: With an MCP Proxy, multiple agents can connect to the same MCP server, necessary for collaboration.

Work in Progress

Current focus is on Goose CLI integration, so you must have it installed. Other agent platforms could integrate easily. The protocol and the MCP server are not tightly coupled to goose, only the npm agent: scripts and the protocol preamble.

New: Behavioral Control with Mermaid Markdown!

Agent protocol expressed as a mermaid chart (example online)
See if your favorite LLM can follow it using the agent:chart script to launch your agent!

How to Run

Configure Goose

  • Run goose configure
  • Choose Add Extension
  • Choose Remote Extension
  • Enter goose-team for name
  • Enter http://localhost:3001/sse for SSE endpoint URI

Run these npm scripts in order:

  • Clone this repo to your local machine and install dependencies
  • npm run build - to build the server from source
  • npm run mcp-proxy to start an SSE proxy to talk to a single instance of the goose-team MCP server
  • npm run inspector to launch the MCP inspector in a browser
    • Click connect, then list_tools, then add_message, then send a message from Human describing a task or project
  • npm run agent to launch an agent that will read the protocol and instructions, assign itself as project coordinator, and create tasks
    • In the inspector, click list_tasks to see tasks
    • Launch additional agents with npm run agent to assign tasks by the Project Coordinator

Troubleshooting

How Well Goose Team Works Depends on the Model

Test if your model will work by running:

  • npm run build
  • npm run mcp-proxy
  • npm run agent:test (launches a simple agent: says hello in different languages, waits 2 seconds, repeats)

Currently, the models that stay in the loop are:

If your configured model stops with "message loop ended," it may not handle tool use well. Try different models.

MCP Tools

Agents

  • register_agent
    Registers a new agent, providing unique ID and random color.

  • list_agents
    Lists all registered agents.

  • agent_leave
    Allows an agent to leave the team.

  • agent_wait
    Allows an agent to wait for a specified number of seconds before next action.

Messages

  • add_message
    Send a new message; stored on the server.

  • recent_messages
    Retrieve the three most recent messages.

  • list_messages
    Retrieve all messages.

  • clear_messages
    Clear all messages.

Tasks

  • add_task
    Add a new task with description.

  • list_tasks
    List all tasks.

  • assign_task
    Assign a task to an agent.

  • complete_task
    Mark a task as complete.

Developer Setup

Install Dependencies

cd /path/to/GooseTeam/
npm install

Build

npm run build

STDIO Server - Proxied

npm run stdio:proxy

Launches an SSE-based MCP proxy on port 3001 at /sse.

SSE Server - Direct

npm run sse:direct

Launches an SSE-wrapped MCP server on port 3001 at /sse.

Inspector

npm run inspector

Runs the Model Context Protocol Inspector UI at http://localhost:5173

  • Set Transport Type to SSE
  • Set URL to http://localhost:3001/sse
  • Click Connect (green light and "Connected" message should appear)
  • Click List Tools

Agent - Protocol as Text

npm run agent:text

Run an agent using natural language instructions (instructions/protocol-as-text.md).

  • First agent assumes Project Coordinator role.
  • Connect to server with Inspector first and send a message from "Human" describing goals.

Agent - Protocol as Chart

npm run agent:chart

Run an agent using Mermaid markdown instructions (instructions/protocol-as-chart.md).

  • First agent assumes Project Coordinator role.
  • Connect to server with Inspector first and send a message from "Human" describing goals.

Agent Wait Test

npm run agent:wait:test

Tests the model's ability to stay in the loop by repeatedly checking messages.

  • Have MCP Proxy running first.
  • If ends with "outgoing message queue empty," model is unsuitable.

Code Formatting and Checks

  • npm run format — runs Prettier to format code.
  • npm run typecheck — runs TypeScript type checks.
  • npm run lint — runs ESLint to check for syntax issues.
  • npm run lint:fix — runs ESLint to fix syntax issues.
  • npm run test — runs unit tests.

Links