microsoft/DebugMCP

What it solves

DebugMCP allows AI coding agents to move beyond reading logs or guessing code behavior by giving them direct, autonomous control over the VS Code debugger. It enables AI assistants to perform operational debugging—such as setting breakpoints and inspecting variables—just as a human developer would.

How it works

It operates as a Model Context Protocol (MCP) server that runs locally within VS Code. The server exposes a set of tools (e.g., start_debugging, add_breakpoint, get_variables_values) that AI agents can call via secure local HTTP communication. It integrates with existing VS Code language debugger extensions to support a wide range of languages without requiring manual launch.json configurations.

Who it’s for

Developers using MCP-compatible AI assistants (such as GitHub Copilot, Cursor, Cline, Windsurf, or Roo Code) who want their AI agents to be able to autonomously debug code across various programming languages.

Highlights

• Comprehensive Debugging Toolset: Includes tools for stepping over/into/out of code, pausing execution, restarting sessions, and evaluating expressions.
• Multi-language Support: Works with Python, JavaScript/TypeScript, Java, C/C++, Go, Rust, PHP, Ruby, and C#/.NET.
• Local and Secure: Runs 100% locally with no external communications, featuring loopback-only binding and automatic secret redaction for variable inspection.
• Test-Aware Debugging: Can launch specific unit tests using the VS Code Testing API to ensure consistent breakpoint hits.
• Zero Config: Automatically generates debug configurations based on file extensions and language.