Code Sentinel

I built Code Sentinel because I kept running into the same problem: AI coding tools are getting really good at changing code, but that's exactly what makes them scary to use on anything real. You paste in a bug, the AI "fixes" it, and now you're left wondering what else it touched, and whether the fix even works. So I designed Code Sentinel around one rule: the AI never changes your code without you seeing it and saying yes first. Here's how it actually works. You describe a bug in plain English, like you would to a teammate. The agent goes and reads through the codebase, searches for relevant files, and comes back with a diagnosis, what it thinks is wrong, why, and which files are involved. From there it proposes an actual fix as a diff, so you can see exactly what would change, before and after, side by side. Nothing happens after that until you click approve. That's the whole point. Once you approve it, Code Sentinel applies the fix and immediately runs your test suite, so you're not just trusting that the fix works, you get a verified or needs-more-work result right away. I think of it as Investigate, Explain, Fix, Approve, Verify. Every step leaves a trail, so by the end you have a diagnosis, a diff, an approval, and a test result tied to one bug report. That's the part I'm most proud of, it's not just an AI writing code, it's an AI that has to show its work and wait for a human before doing anything. This is built for developers who want AI help debugging without handing over control of their codebase, which is honestly most developers I know. Under the hood it's a React and Vite frontend, a Node/Express backend, and Groq's API doing the diagnosis and fix generation, with Node's own test runner closing the loop.