Python Code Execution MCP Server

A Model Context Protocol (MCP) server that allows AI assistants to execute Python code with persistent state across multiple calls.

Features

• Python Code Execution: Execute Python code with your installed Python interpreter. No sandbox!!!
• Persistent State: Variables and functions defined in previous executions remain available in subsequent calls
• Environment Management: Tools to reset environment and inspect defined variables

Installation

1. Install dependencies in your Python environment:

pip install fastmcp

2. Clone this repository or download server.py.

Usage

Adding the Server to any MCP-compatible chat client

{
"python-interpreter": {
	"type": "stdio",
	"command": "path_to_your_python_exe",
	"args": [
	"path_to_the_server.py"
	]
}
}

Available Tools

run_python_code

Execute Python code and return detailed results. Variables and functions defined in previous executions will remain available in subsequent calls.

Parameters:

• code (string): The Python code to execute

Returns:

• Execution status (success/failure)
• Return code
• Standard output
• Standard error
• Execution time

reset_python_environment

Reset the Python execution environment, clearing all variables and imports.

Returns:

• Confirmation message

list_defined_variables

List all currently defined variables and functions in the Python environment.

Returns:

• List of defined variables and functions with their types and values

Dependencies

• fastmcp: MCP server framework
• Python 3.6+: Required for the server runtime

Technical Notes

• Persistent State: Uses a global namespace to maintain variables and functions across executions
• No Subprocess: Executes code directly in the server process for better performance and state persistence
• UTF-8 Support: Handles Unicode characters properly without additional encoding setup
• Error Handling: Captures both stdout/stderr and Python exceptions with full stack traces
• Working Directory: Preserves the server's current working directory
• Standard Library: All standard library modules are available
• Third-party Packages: Additional packages must be installed in the server's Python environment

Security Warning

⚠️ This server executes arbitrary Python code without sandboxing. Only use in trusted environments.

License

MIT

Support

For issues and questions:

• Create an issue in the repository
• Check the FastMCP documentation for MCP-related questions