rish-e/debuggai
The universal verification layer for AI-generated software
Platform-specific configuration:
{
"mcpServers": {
"debuggai": {
"command": "npx",
"args": [
"-y",
"debuggai"
]
}
}
}Add the config above to .claude/settings.json under the mcpServers key.
<p align="center"> <h1 align="center">DebuggAI</h1> <p align="center">The universal verification layer for AI-generated software.</p> </p>
<p align="center"> <a href="https://github.com/rish-e/debuggai/blob/main/LICENSE"></a> </p>
---
AI-generated code has 1.7x more bugs than human-written code. DebuggAI catches them — hallucinated APIs, security vulnerabilities, performance anti-patterns, and intent mismatches — before they reach production.
---
There are two ways to use DebuggAI. Pick whichever fits your workflow.
Use DebuggAI without leaving your AI coding workflow. Just talk naturally.
One-time setup:
pip install debuggai
debuggai setupThen restart Claude Code (or Cursor).
That's it. Now just ask Claude:
Claude will use DebuggAI's tools automatically. No commands to memorize — just describe what you want.
Run scans directly from your terminal.
cd ~/my-project
debuggai scan --no-llm # Scan everything (no API key needed)
debuggai scan --file src/app.py # Scan specific file
debuggai scan --diff HEAD~1 # Scan what changed since last commit
debuggai scan --staged # Scan staged changes before committing
debuggai verify --intent "add OAuth" # Verify code matches what you asked AI to build
debuggai fix # Generate fixes for detected issues
debuggai fix --apply # Auto-apply high-confidence fixes
debuggaiLoading reviews...