MCP Router

What is MCP Router?

MCP Router is a Python package for interacting with Model Context Protocol (MCP) servers with OpenRouter LLM integration. It provides a backend for managing MCP servers, executing agent-based tasks, and orchestrating complex workflows using the Upsonic framework.

How to use MCP Router?

Install via git clone and pip, then import as a Python module or run the CLI with python -m mcp_router.cli.cli --help. Use the ServerManager class to manage servers and the OpenRouter integration to query LLMs.

Key features of MCP Router

• MCP server management (add, edit, remove)
• OpenRouter integration for AI‑assisted tasks
• Upsonic integration for multi‑step workflows
• Intelligent task analysis to select tools
• Standardized MCP protocol support
• REST API endpoints for functionality exposure

Use cases of MCP Router

• Orchestrate complex multi‑step workflows using Upsonic and MCP tools
• Query multiple LLM providers via a single OpenRouter API
• Automatically determine which MCP tools are needed for a given task
• Manage and interact with local MCP servers from a Python application

FAQ from MCP Router

What are the prerequisites for using MCP Router?

Python 3.8+ is required. Docker is optional but recommended for running MCP servers.

How does MCP Router manage MCP servers?

It reads MCP server configuration from ~/.cursor/mcp.json, and provides capabilities to start, stop, and communicate with those servers.

Does MCP Router support authentication or transport layers?

The README mentions a REST API framework and MCP protocol support but does not specify particular authentication or transport details beyond those.

What data does MCP Router store and where?

Server configuration is read from ~/.cursor/mcp.json; no other data storage locations are mentioned.

Can I use MCP Router without Docker?

Yes, Docker is optional. MCP servers can be run without it, though Docker is recommended for easier setup.