env-doctor

Overview

Env-Doctor is a comprehensive tool that diagnoses and fixes GPU environment compatibility issues, particularly focusing on CUDA version mismatches between NVIDIA drivers, system toolkits, cuDNN, and Python AI libraries. The MCP server exposes 11 diagnostic tools to AI assistants like Claude Desktop, enabling them to check GPU environments, validate Dockerfiles, determine model compatibility, and provide remediation solutions. It works across local systems, Docker containers, and CI/CD pipelines with support for WSL2 and multiple cloud platforms.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• check

  Perform full environment compatibility diagnosis between GPU driver, CUDA toolkit, cuDNN, and Python libraries.

  Try:Check my GPU environment for compatibility issuesTry:Diagnose why PyTorch is failing with CUDA errors
• python-compat

  Check Python version compatibility with AI libraries and detect dependency cascade issues.

  Try:Check if Python 3.13 is compatible with my installed librariesTry:Show Python version conflicts with TensorFlow and PyTorch
• cuda-install

  Install CUDA Toolkit with automatic platform detection and verification.

  Try:Install CUDA toolkit for my systemTry:Show me what CUDA installation commands would run
• install

  Get safe installation commands for AI libraries that match your CUDA version.

  Try:Get the correct PyTorch installation command for my systemTry:Install flash-attn with proper CUDA compatibility
• model

  Check if AI models fit in GPU memory and get cloud GPU recommendations.

  Try:Check if Llama 3 70B will fit on my GPUTry:Show cloud GPU recommendations for a model
• dockerfile

  Validate Dockerfiles for GPU configuration and Python library compatibility.

  Try:Validate this Dockerfile for GPU issuesTry:Check if my Dockerfile has the right CUDA image
• docker-compose

  Validate docker-compose.yml files for GPU configuration issues.

  Try:Check my docker-compose.yml for GPU configuration problemsTry:Validate docker-compose setup for PyTorch container
• cuda-info

  Provide detailed analysis of CUDA toolkit installation and configuration.

  Try:Show detailed information about my CUDA installationTry:Detect multiple CUDA installations on my system
• cudnn-info

  Analyze cuDNN library installation and compatibility with CUDA.

  Try:Check if cuDNN is properly installedTry:Show cuDNN version compatibility with my CUDA
• scan

  Scan code for deprecated CUDA-related imports.

  Try:Scan my code for deprecated PyTorch importsTry:Find CUDA API calls that need updating
• debug

  Provide verbose diagnostic output for troubleshooting environment issues.

  Try:Show detailed debug information for my GPU environmentTry:Get verbose output for CUDA configuration problems
• init

  Generate CI/CD workflow files for GPU environment monitoring.

  Try:Create a GitHub Actions workflow for GPU environment checksTry:Generate CI configuration for monitoring CUDA compatibility

Comparable tools

Installation

pip install env-doctor

{
  "mcpServers": {
    "env-doctor": {
      "command": "env-doctor-mcp"
    }
  }
}

FAQ

What problems does Env-Doctor solve?

It diagnoses and fixes GPU/CUDA compatibility issues, particularly version mismatches between NVIDIA drivers, CUDA toolkit, cuDNN, and Python AI libraries like PyTorch and TensorFlow.

How does the MCP server integrate with AI assistants?

It exposes 11 diagnostic tools to AI assistants through the Model Context Protocol, enabling them to check GPU environments, validate configurations, and suggest fixes.

Compare env-doctor with