LLDB-MCP is an MCP server that connects Claude’s Model Context Protocol to the LLDB debugger, enabling AI-driven debugging workflows. It lets Claude start, manage, and interact with LLDB sessions directly, translating natural language prompts into debugging actions. The server supports creating and managing multiple LLDB sessions, loading executables and attaching to running processes, and loading core dump files for post-mortem analysis. Users can issue arbitrary LLDB commands, and have fine-grained control over program execution (run, continue, step, next, finish). It provides memory examination and disassembly, as well as thread and stack frame inspection, to inspect program state. The system includes session management (list, terminate), breakpoint and watchpoint support, and a variety of inspection commands (backtrace, print, examine, etc.). The README also outlines example Claude prompts and a configuration snippet to register the MCP server in the Claude desktop app. Debug tips suggest enabling a debug mode for verbose logs and note that sessions are cleaned up on shutdown. Requirements include Python 3.7+, LLDB installed, and Claude MCP support.