Killer Bug AI Debugger

Let your AI assistant debug your code automatically for accurate runtime context

An MCP-powered debugging extension that gives VS Code Copilot, Cursor AI, Claude Desktop, ChatGPT, and other AI assistants complete control over VS Code's debugger. AI can now set breakpoints, step through code, inspect variables, and find bugs—all without you touching the keyboard.

🎬 See It In Action

Killer Bug AI Debugger Demo

⚡ Quick Start

Install the extension from VS Code marketplace
Start the Debugger (for each project):
- Click the "Killer Bug: Ready (Click to Start)" button in the status bar (bottom left)
- Or press Ctrl+Shift+P (or Cmd+Shift+P on Mac) and run: Killer Bug: Start AI Debugger
Configure (first time only):
- The extension auto-detects your IDE (VS Code or Cursor)
- Choose the port for the MCP server (usually 3100)
Connect your AI:
- VS Code: Open mcp.json → Click "Start" icon on top of "killer-bug-" in servers
- Cursor: Go to Settings → Enable MCP Server "killer-bug-"
Start debugging: Chat with your AI: "Debug this code and find the bug"

That's it! Your AI debugger is ready.

💡 Tip: The extension auto-starts when you open a configured project (can be disabled in settings). For unconfigured projects, just click the status bar button to get started!

🎯 What It Does

🎬 Live Debugging

Your AI can:

Start debugging any file (Python, JavaScript, TypeScript, etc.)
Set breakpoints and conditional breakpoints
Step through code line by line
Inspect variables and the call stack
Evaluate expressions in real-time

[GIF: AI Setting Breakpoint and Stepping Through Code]

🔗 Remote Debugging

Attach to running servers (FastAPI, Node.js, Flask, etc.)
Debug over the network or in containers
Use your existing launch.json configurations
Perfect for production debugging

[GIF: AI Attaching to Running FastAPI Server and Inspecting Variables]

🧠 AI-Powered Bug Hunting

AI automatically explores your code
Sets strategic breakpoints
Finds and explains bugs
Suggests fixes

[GIF: AI Finding and Explaining a Bug]

📚 How It Works

Your AI → MCP Protocol → Killer Bug → VS Code Debugger → Your Code

The extension runs an HTTP server that implements the Model Context Protocol. Your AI connects and controls the debugger through 17 specialized tools.

🛠️ Features & Tools

Session Control

Start debugging any file (auto-detects language)
Stop debug session
Check execution state
List and use existing launch.json configurations
Attach to running processes

Breakpoints

Set breakpoints with conditions
Remove breakpoints
List all breakpoints

Execution

Continue to next breakpoint
Step over/into/out of functions
Pause execution

Inspection

View call stack with function names & line numbers
Inspect variables in current scope
Evaluate expressions safely

🚀 Installation & Setup

Marketplace (Recommended)

Open VS Code → Extensions
Search Killer Bug AI Debugger
Click Install

Development

git clone <repo>
cd vscode-debug-mcp
npm install
npm run build
# Press F5 to launch

Configure for Your AI

Starting the Debugger:

Click "Killer Bug: Ready (Click to Start)" in the status bar, OR
Press Ctrl+Shift+P / Cmd+Shift+P and run: Killer Bug: Start AI Debugger
If not configured yet, select your IDE (VS Code or Cursor) and choose a port
Done! The MCP server is now running

Connecting your AI client:

Go to your AI client's MCP settings
Add the Killer Bug server (if not auto-detected)
Start debugging!

Or manual configuration for your MCP client:

{
  "mcpServers": {
    "killer-bug-debugger": {
      "type": "http",
      "url": "http://localhost:3100/mcp"
    }
  }
}

Settings

Auto-Start Configuration:

Setting: killerBug.autoStart (default: enabled)
When enabled: MCP server automatically starts when you open a project that's already configured
When disabled: Server won't start automatically; click the status bar button to start
Configure per-workspace in VS Code Settings under "Killer Bug AI Debugger"

💬 Usage Examples

Basic Debugging:

"Debug the file app.py and set a breakpoint at line 42"

Step Through Code:

"Step through the calculate_total function and show me the variables"

Inspect State:

"What's the current call stack? Show me the value of user"

Remote Debugging:

"Attach to my FastAPI server running on port 5678 and debug the /api/users endpoint"

Using Configurations:

"Start debugging using my 'FastAPI Debug' configuration from launch.json"

Supported Languages

Automatically debugs:

Python (.py)
JavaScript/TypeScript (.js, .ts, .jsx, .tsx)
And any language with VS Code debug adapter support

❓ FAQ

Q: Is my code safe? A: Yes. The debugger only reads variables and evaluates safe expressions. No code modification without explicit AI request.

Q: Works with any AI? A: Yes! Works with VS Code Copilot, Cursor AI, Claude Desktop, ChatGPT, and any AI/LLM that supports MCP protocol.

Q: What about privacy? A: The server runs locally. All debugging happens on your machine. No data sent anywhere.

Q: Can I use it for production? A: Yes! Remote debugging is perfect for debugging live servers safely.

🐛 Troubleshooting

Extension not starting?

Check Output → Extension Host for errors
Ensure port 3100 is available

AI can't connect?

Run: curl http://localhost:3100/health
Verify MCP client configuration

Debugging tools not working?

Start a debug session first
Check Debug Console for logs

📖 More Information

� Contributing

This project is open source and we'd love your help!

Have ideas for new features? Found a bug? Want to improve the code?

Fork the repository - https://github.com/fellowabhi/Killer-bug
Create a feature branch - git checkout -b feature/your-feature
Make your changes and test thoroughly
Submit a Pull Request - We review all PRs and love community contributions!

All contributions are welcome, from bug fixes to new debugging features to documentation improvements.

�📄 License & Open Source

GPL-3.0-or-later - Free and Open Source Software

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

What this means:

✅ Source code is publicly available on GitHub
✅ You can modify and redistribute it
✅ Any modifications must also be open source
✅ No warranty, but completely transparent

See LICENSE file for full details.

Source Code: https://github.com/fellowabhi/Killer-bug

Happy debugging! 🚀

Killer Bug - Debugger MCP Server

fellowabhi