This project provides an integration for LLDB, enabling AI assistants like Claude to interact with debugging sessions via the Model-Context Protocol (MCP). It targets developers seeking to leverage natural language for debugging complex codebases, offering a bridge between AI models and low-level debugger functionality.

How It Works

The integration comprises two core components: server.py (an MCP server) and lldb_plugin.py (an LLDB plugin). The plugin exposes LLDB's capabilities through JSON-RPC, which the server then translates into MCP messages for AI clients. This architecture allows AI assistants to issue commands like setting breakpoints, stepping through code, and inspecting variables, receiving structured feedback from LLDB.

Quick Start & Requirements

• Installation: Clone the repository, then pip install "fastmcp>=1.2.0" httpx or uv.
• Prerequisites: Python 3.10+, LLDB with Python bindings, fastmcp package.
• Setup: Configure Claude for Desktop's claude_desktop_config.json or add command script import /path/to/lldb_plugin.py to ~/.lldbinit and run (lldb) mcp start.
• Docs: https://github.com/ant4g0nist/lldb-mcp

Highlighted Details

• Exposes 15+ LLDB commands via MCP, including create_target, set_breakpoint, get_backtrace, and evaluate_expression.
• Supports direct LLDB integration or use with Claude for Desktop.
• Allows execution of arbitrary LLDB commands (run_lldb_command).
• Provides functionality to read memory and get disassembly.

Maintenance & Community

• Contributions are welcome via pull requests and issues.
• Mentions credits to voltron, lldb, chisel, gef, pixd, hexdump, idaref, and ida-pro-mcp.

Licensing & Compatibility

• Licensed under the Apache License.
• Compatible with commercial use and closed-source linking.

Limitations & Caveats

The project is marked with a TODO for updating instruction manuals and adding more test cases, suggesting it may still be under active development or refinement.