env-doctor

Env-Doctor is an MCP server that exposes a suite of GPU and Python environment diagnostics to AI assistants via the Model Context Protocol. It wraps diagnostic capabilities such as environment checks, Python compatibility, CUDA and cuDNN details, and installation guidance, and makes them queryable by AI agents like Claude Desktop or Claude Code. With the MCP endpoint, you can ask for comprehensive environment diagnoses, specific component checks, CUDA toolkit details, safe installation commands for libraries, and model compatibility assessments, all encoded in a structured JSON output suitable for automation and reasoning by AI assistants.

Prerequisites:

• Python 3.10+ (as indicated by Env-Doctor’s documentation)
• Access to a Python environment with network access to install packages

Installation steps:

1. Ensure Python 3.10+ is installed and available on your PATH.
2. Upgrade pip:

   python -m pip install --upgrade pip
   

3. Install the Env-Doctor MCP server package (which provides env-doctor and the MCP server wrapper):

   pip install env-doctor
   

4. Run the MCP server for integration with MCP clients:

   env-doctor-mcp
   

Notes:

• If the command env-doctor-mcp is not on your PATH, locate the installed console script and run it directly, or invoke via Python:

  python -m env_doctor.mcp_server
  

• You can configure the MCP server in your AI assistant’s config by referencing the command above (see README integration guide for Claude Desktop).

Tips and common issues:

• Ensure you are using a Python environment compatible with Env-Doctor (Python 3.10+).
• If you encounter environment or GPU-related checks failing, verify that CUDA drivers and toolkit installations on the host match the expectations of your AI libraries as diagnosed by Env-Doctor.
• The MCP server outputs JSON-structured diagnostics suitable for automated processing by AI assistants; use those outputs to drive model-based decision making in workflows.
• When integrating with Claude Desktop, keep the MCP server command consistent with the config snippet provided in the integration guide to ensure automatic tool availability.
• No mandatory environment variables are required for a basic run; advanced users may supply additional env vars if their deployment environment requires custom paths or proxies.