Skip to content

facebookexperimental/dapper

Repository files navigation

Dapper

Dapper is a tool that provides multiple clients with simultaneous access to debugging sessions that is uniform across different programming languages and environments.

The goal of the project is to enable agentic AI debugging through a universal MCP server for debugging. It enables access to debugging sessions in various settings:

  • Autonomous. Agents can drive debugging sessions without human involvement. Imagine inspecting and summarizing findings from analyzing core dumps.
  • Collaborative. An AI agent jumps into your existing debugging session. Perhaps it has new insights into why that thread hangs.
  • Introspective. Follow along as an AI agent drives a debugging session in the IDE, all through the familiar debugging interface of VS Code.

How It Works

Dapper is a debug proxy server that sits between DAP clients (like VS Code) and language-specific debug adapters (like LLDB, debugpy). It works with any DAP server (lldb-dap, debugpy, dlv, etc.) without requiring modifications to existing components.

Why a proxy? DAP does not support multi-client interactions with a debugging session out of the box. The proxy enables us to offer AI agents insights and control of existing debugging sessions.

License

MIT


dapper /'dæpər/ n. 1. In computing, a system or entity that utilizes the Debug Adapter Protocol (DAP). 2. Dutch. brave.

About

Give AI agents access to debugging sessions through the Model Context Protocol (MCP) that is uniform and consistent across different programming languages and environments.

Resources

License

Code of conduct

Contributing

Security policy

Stars

7 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages