MCP Connector

What is MCP Connector?

MCP Connector is a lightweight server that runs and manages multiple Model Context Protocol (MCP) servers, designed to integrate with TypingMind. It enables running MCP servers on a local computer or remote server and connecting custom AI models or tools to TypingMind via a simple REST API.

How to use MCP Connector?

Run instantly with npx @typingmind/mcp@latest <auth-token> (no install required) or provide the token via the MCP_AUTH_TOKEN environment variable. Keep the process running while using TypingMind. For Docker, build an image and run with mapped ports. Optionally set TIMEOUT, PORT, or HTTPS environment variables.

Key features of MCP Connector

• Manages multiple MCP servers simultaneously
• REST API with endpoints for clients and tools
• Health check endpoint (/ping)
• Tool call timeout configuration (default 60 seconds)
• HTTPS support via certificate and key files
• Docker support for containerized deployment

Use cases of MCP Connector

• Connect custom AI models or tools to TypingMind
• Run MCP servers locally or on a remote server
• Proxy HTTP requests through the /mcp-connect endpoint
• Manage and restart individual MCP clients via API

FAQ from MCP Connector

How do I get the authentication token?

The authentication token is provided by TypingMind. Follow the instructions at www.typingmind.com/mcp to obtain it.

What are the runtime requirements?

Node.js version 14 or later is required. Optionally, Docker can be used for containerized deployment.

Can MCP Connector run on a remote server?

Yes. Run it using npx or Docker on a remote server, optionally with a process manager like pm2 for persistence.

What is the default tool call timeout?

The default timeout is 60 seconds (60000ms). It can be customized via the TIMEOUT environment variable.

Which ports does the server use?

The server defaults to port 50880 or 50881. You can specify a different port using the PORT environment variable.