Gooseteam

🚀 GooseTeam Project Introduction

This is an application project for creating a collaborative AI team. By leveraging large language models (LLMs), it enables multitasking, process automation, and complex decision - making processes.

🚀 Quick Start

1. Launch MCP agents:

   npm run agent:text # Launch the agent using the text protocol
   npm run agent:chart # Launch the agent using the chart protocol
   npm run agent:wait:test # Test the agent's wait - loop ability
   

2. Run the inspector tool:

   npm run inspector
   

   • Open the Inspector UI: http://localhost:5173
   • Set the transmission type to SSE
   • Set the URL to http://localhost:3001/sse
   • Click the connect button

3. Server startup commands:

   npm run sse:direct # Start the direct SSE server
   npm run sse:proxy # Start the proxy SSE server
   

✨ Features

• Agent Launch: Multiple AI agents can be launched through different protocols.
• Server Configuration: Supports various server configurations, including direct SSE and proxy modes.
• Tool Inspection: Provides an inspector tool for monitoring and managing agents.
• Process Testing: Includes wait - loop tests to evaluate the model's continuous interaction ability.

💻 Usage Examples

Basic Usage

Launching Agents

# Launch agents with different protocols
npm run agent:text # Use text protocol
npm run agent:chart # Use chart protocol
npm run agent:wait:test # Test wait - loop ability

Running the Inspector

npm run inspector

• Open the Inspector UI at http://localhost:5173, set the transmission type to SSE, the URL to http://localhost:3001/sse, and click the connect button.

Starting the Server

npm run sse:direct # Start direct SSE server
npm run sse:proxy # Start proxy SSE server

Advanced Usage

The advanced usage mainly involves the combination of different commands according to specific scenarios, such as first starting the server and then launching agents for testing, etc.

📚 Documentation

Inspector Tool

• Run npm run inspector to start the inspector tool.
• The UI interface offers multiple functions:
  • Connect to the Server: Set the correct transmission type and URL, then click the connect button.
  • List Tools: View currently available tools and their statuses.
  • Monitor Resources: Monitor the agent's interaction in real - time.

Agent Protocols

Text Protocol (agent:text)

• Launch the agent using the text instructions in instructions/protocol - as - text.md.
• The first agent automatically assumes the role of project coordinator.

Chart Protocol (agent:chart)

• Launch the agent using the chart instructions in instructions/protocol - as - chart.md.
• This protocol uses the Mermaid format, suitable for users with a technical background.

Wait - Loop Test (agent:wait:test)

• Test whether the agent can maintain a loop state for message checking.
• If a queue - empty error occurs, the model is not suitable for team collaboration.

Formatting and Code Standards

npm run format # Format code using Prettier
npm run typecheck # Run TypeScript type checking
npm run lint # Perform ESLint code inspection
npm run lint:fix # Automatically fix code issues

Unit Testing

npm test # Run the unit test suite

⚠️ Important Note

• Ensure the server is started before running the agents.
• It is recommended to use verified LLM models, such as those verified by the GooseTeam.
• Run the project in an environment that supports SSE.

💡 Usage Tip

This project aims to achieve complex collaborative tasks through AI agents, suitable for application scenarios requiring multi - step processing and decision - making.

Links and Resources

• MCP Specification: The complete Model Context Protocol specification can be found here.
• Server Reference: A simple yet modular GitHub example is used as a development reference.
• Inspiration Source: Based on Aaron Goldsmith's Gist and his presentation video.