medical

Medical MCP is a local, data-sovereign server that compiles a range of authoritative medical data sources into a private API for your AI workflows. It exposes tools for drug information, health statistics, and medical literature, including PubMed, RxNorm, FDA data, WHO statistics, Google Scholar results, and pediatric guidelines. You can query these capabilities directly via MCP-compatible clients, or configure clients like Claude Desktop to route queries through the local server. The server emphasizes privacy and zero API keys, ensuring everything runs offline on your machine. Typical use cases include researching drug details, retrieving global health indicators, and locating relevant medical literature without leaving your environment.

Prerequisites: Node.js 18+ and npm. Ensure you have a working development environment and network access for initial install if you plan to build from source.

1. Install globally via npm (recommended):

npm install -g medical-mcp

2. If you prefer to install from source:

# Clone the repository
git clone https://github.com/JamesANZ/medical-mcp.git
cd medical-mcp

# Install dependencies and build
npm install
npm run build

3. Run the server locally:

npm start

4. Claude Desktop configuration (example):

# In claude_desktop_config.json, add a medical-mcp entry
"mcpServers": {
  "medical-mcp": {
    "command": "node",
    "args": ["/absolute/path/to/medical-mcp/build/index.js"]
  }
}

If you prefer Cursor on-click install, use the provided one-click install link in the repository README.

Notes and tips:

• The server runs locally and does not send data to external services unless explicitly queried.
• You can tune caching with environment variables listed in the README. Default TTLs cover FDA, PubMed, WHO, RxNorm, and pediatric resources.
• If you modify data sources or rebuild, ensure you restart the MCP server to pick up changes.
• The npm package name is medical-mcp; use the node-based invocation path shown in Claude Desktop configuration for production setups.
• For troubleshooting, check cache settings and ensure your environment variables are exported to the process hosting the MCP server.