Overview
What is Simple MCP Server?
Simple MCP Server is a minimal implementation of the Model Context Protocol (MCP) server built with FastMCP. It demonstrates how to create an MCP server that clients like Windsurf IDE and Claude can connect to, providing example tools and dual transport support.
How to use Simple MCP Server?
Clone the repository and run uv sync to install dependencies. Then execute python run_server.py to start both HTTP (port 9876) and stdio transports simultaneously. Use --http-only or --stdio-only to run a single transport. Configure Windsurf or Claude Desktop to connect as shown in the README.
Key features of Simple MCP Server
- Dual transport support (HTTP and stdio)
- Token-based authentication for HTTP transport
- Example tools: hello_world, get_version, system_info
- Compatible with Windsurf IDE and Claude Desktop
- Minimal, easy-to-extend codebase
Use cases of Simple MCP Server
- Learning how to build an MCP server with FastMCP
- Prototyping tools for AI assistants like Claude
- Testing MCP client connectivity and authentication flows
- Serving as a starter template for custom tool implementations
FAQ from Simple MCP Server
What is Simple MCP Server’s purpose?
It is a minimal example of an MCP server using FastMCP, intended to show how to create tools that LLM clients like Windsurf IDE and Claude can call.
What are the prerequisites?
Python 3.11 or higher, FastMCP, and uv. Install dependencies with uv sync after cloning the repository.
How do I run Simple MCP Server?
Run python run_server.py from the project directory. For HTTP only, add --http-only; for stdio only, add --stdio-only. The HTTP server listens on port 9876.
How do I connect Windsurf IDE to Simple MCP Server?
Use the configuration shown in the README: a command with npx mcp-remote pointing to http://localhost:9876/ with the header Authorization: Bearer SIMPLE_MCP_SERVER.
How do I connect Claude Desktop to Simple MCP Server?
Add an MCP server entry in Claude’s configuration with command uv run --project <path> python <path>/run_server.py --stdio-only, replacing <path> with the full path to the project.
