pexpect-mcp

An MCP (Model Context Protocol) server that provides remote pexpect session control for debugging and process interaction.

Overview

This MCP server enables AI assistants to execute Python code with pexpect functionality, allowing for interactive debugging sessions with tools like LLDB, GDB, and other command-line utilities that require programmatic interaction.

In some sense this is less of a pexpect MCP as one that is just maintaining a stateful Python session.

Installation

uv tool install git+https://github.com/mitsuhiko/pexpect-mcp

Usage

As an MCP Server

Add to your Claude Code configuration:

{
  "mcpServers": {
    "pexpect": {
      "command": "pexpect-mcp"
    }
  }
}

Tool Usage

The server provides a single tool: pexpect_tool

Parameters:

• code (string): Python code to execute with pexpect support
• timeout (optional int): Timeout in seconds (default: 30)

Example Usage:

# Start a debugging session
child = pexpect.spawn('lldb ./my-program')
child.expect('(lldb)')

# Run the program
child.sendline('run')
child.expect('(lldb)')
print(child.before.decode())

# Get backtrace
child.sendline('bt')
child.expect('(lldb)')
print(child.before.decode())

Demo

The repository includes a demo with a buggy C program (demo-buggy.c) that can be debugged using LLDB through the pexpect interface. This demonstrates the server’s capability for interactive debugging sessions.

The program `./demo-buggy` crashes when executed. Use LLDB to:

- Start the program under the debugger
- Identify where and why it crashes
- Examine variables, memory, and call stack
- Report the root cause of the crash

Requirements

• Python ≥ 3.12.1
• pexpect ≥ 4.9.0
• mcp ≥ 1.13.0

License

See LICENSE file for details.