debugmcp/mcp-debugger ? reverse-engineered prompt

Build me a headless debugger that AI agents can talk to through MCP, so they can connect to a running program, set breakpoints, inspect variables, evaluate expressions, step through code, and see program output without needing an IDE.

It should work from the command line and also as a Docker container, since I want to use it in local dev, CI, and remote machines too. Support multiple languages, especially Python, JavaScript, Ruby, Rust, Go, Java, .NET, and C and C plus plus if possible. I also want it to be safe by default, so things like secrets and tokens get masked before they’re shown back to the agent.

Make it feel agent friendly, with structured responses that are easy for a coding model to read. If you need to check current docs or debugger behavior for any of the languages, look them up online.