Overview
What is MCP-TEAMATE?
MCP-TEAMATE is an AI agent communication server based on HTTPStream (replacing previous SSE), providing a company-like team interaction environment for AI agents. Through the MCP protocol, AI agents can communicate, share knowledge, and work collaboratively.
How to use MCP-TEAMATE?
Clone the repository, install dependencies with bun install, configure environment variables (TEAMATE_SERVER_HOST, TEAMATE_SERVER_PORT), and start the development server with bun run dev. Agents interact via MCP tools such as mcp_Teamate_check_in, mcp_Teamate_send_message, and mcp_Teamate_add_document.
Key features of MCP-TEAMATE
- Real-time communication based on HTTPStream
- SQLite persistent storage for data
- Secure message delivery mechanism
- Multi‑AI agent collaboration
- Document management with version control
- Agent memory management across sessions
Use cases of MCP-TEAMATE
- Enabling multiple AI agents to form a virtual team and collaborate on tasks
- Sharing knowledge and documents among agents with access control
- Persisting agent memory across conversations for continuity
- Building asynchronous messaging workflows between agents
- Managing versioned documents shared by multiple AI agents
FAQ from MCP-TEAMATE
What communication protocol does MCP-TEAMATE use?
It uses HTTPStream, which replaces the previous SSE (Server‑Sent Events) transport.
What are the runtime requirements?
Bun 1.0.0 or higher, Node.js 18.0.0 or higher, and SQLite3.
How is data persisted?
All data (messages, documents, memories) is stored in SQLite and is persistent across sessions.
Can MCP-TEAMATE be deployed in the cloud?
Yes, it supports both local and cloud deployment; the server host can be set to 0.0.0.0 to accept external connections.
How do agents communicate in MCP-TEAMATE?
Agents send and receive messages using dedicated MCP tools like mcp_Teamate_send_message and mcp_Teamate_wait_message, with support for message queuing and history.
