Dap Mcp

Overview

dap‑mcp is a Model Context Protocol (MCP) server that bridges the gap between large language models and traditional debugging workflows. By exposing a DAP‑compatible interface through MCP, it lets AI assistants such as Claude control debuggers, manipulate breakpoints, step through code, and inspect execution state—all within the conversational context of a language model. This solves the problem of limited debugging visibility for AI assistants, enabling them to provide deeper, context‑aware guidance during development.

The server implements the full Debug Adapter Protocol stack while delegating actual debugging to well‑established backends like debugpy and lldb. It offers a rich set of tools—launching the debuggee, setting or removing breakpoints, stepping, evaluating expressions, switching stack frames, and viewing source snippets. These tools are wrapped in MCP messages that preserve context across interactions, allowing the assistant to remember where it left off and to continue a debugging session seamlessly. The XML‑rendered output from each tool is designed for easy integration with MCP clients, ensuring that responses can be displayed or further processed by the AI.

Key capabilities include:

• Configurable DAP backends through a JSON file that specifies the debugger executable, arguments, source directories, and other launch options.
• Fine‑grained breakpoint control with optional conditions and source resolution via .
• Execution flow manipulation (continue, step‑in/out/next) that can be invoked directly from the conversation.
• Contextual evaluation of expressions and stack frame navigation, enabling the assistant to answer “What is this variable?” or “Why did we hit this breakpoint?”.
• Source code insight through , which presents a contextual snippet that the assistant can reference.

Typical use cases involve developers who want their AI partner to act as a live debugging companion. In an IDE integration, the assistant can suggest breakpoint placements, explain why a particular line is hit, or automatically step through failing tests. In CI/CD pipelines, the server can be queried to reproduce failures and report stack traces back to the model for diagnostic suggestions. Because the server adheres strictly to MCP, it can be plugged into any AI workflow that supports the protocol, from web‑based chatbots to command‑line tools.

What sets dap‑mcp apart is its seamless MCP integration combined with a fully featured DAP implementation. Developers can extend the server to support additional DAP backends simply by adding a new configuration class, thanks to the shared base. This modularity means that as new debuggers emerge, the server can evolve without rewriting its core logic. The result is a powerful, extensible bridge that lets AI assistants dive into the heart of your codebase and provide actionable debugging insights.