MCP Python Executor

A Model Context Protocol (MCP) server for executing Python code and managing Python packages.

Features

• Execute Python code with safety constraints

• Install and manage Python packages

• Pre-configure commonly used packages

• Resource monitoring and limits

• Health checks and metrics

• Structured logging

Related MCP server: Lab Virtual MCP Server

Configuration

The server can be configured through environment variables in the MCP settings:

{
  "mcpServers": {
    "mcp-python-executor": {
      "command": "node",
      "args": ["path/to/python-executor/build/index.js"],
      "env": {
        "PREINSTALLED_PACKAGES": "numpy pandas matplotlib scikit-learn",
        "MAX_MEMORY_MB": "512",
        "EXECUTION_TIMEOUT_MS": "30000",
        "MAX_CONCURRENT_EXECUTIONS": "5",
        "LOG_LEVEL": "info",
        "LOG_FORMAT": "json"
      }
    }
  }
}

Environment Variables

• PREINSTALLED_PACKAGES: Space-separated list of Python packages to install on startup

• MAX_MEMORY_MB: Maximum memory limit per execution (default: 512)

• EXECUTION_TIMEOUT_MS: Maximum execution time in milliseconds (default: 30000)

• MAX_CONCURRENT_EXECUTIONS: Maximum number of concurrent executions (default: 5)

• LOG_LEVEL: Logging level (debug|info|error, default: info)

• LOG_FORMAT: Log format (json|text, default: json)

Available Tools

1. execute_python

Execute Python code and return the results.

interface ExecutePythonArgs {
  code?: string;          // Python code to execute (inline)
  scriptPath?: string;    // Path to existing Python script file (alternative to code)
  inputData?: string[];   // Optional input data
}

Examples:

// Example with inline code
{
  "code": "print('Hello, World!!')\nfor i in range(3): print(i)",
  "inputData": ["optional", "input", "data"]
}

// Example with script path
{
  "scriptPath": "/path/to/your_script.py",
  "inputData": ["optional", "input", "data"]
}

2. install_packages

Install Python packages.

interface InstallPackagesArgs {
  packages: string[];
}

Example: