Most AI debugging tools are chatbots you paste stack traces into. Corvin is different, it's a CLI agent that plugs into your running services over WebSocket, watches real requests, and traces failures across service boundaries: contract mismatches between services, missing config for an edge-case input, race conditions between timeouts and async completions. Built for multi-service backends where the bug isn't in one file, it's in the seam between two services.
Hey Product Hunt 👋
I built Corvin because I was tired of AI debugging tools that only see the code you paste in front of them. Real bugs, especially in multi-service backends, live in the gaps: a `fullName` field one service sends and another expects as `firstName`/`lastName`, a tax config missing a country code, a timeout that fires 1 second before the async job it's racing against completes.
Corvin is a CLI agent that connects to your actual running services over WebSocket, watches a failing request end-to-end, and traces the fault across service boundaries, not just within one file.
How it works:
1. Run the CLI, point it at your local stack
2. Trigger the failing scenario
3. Corvin walks the real request path across services and surfaces root cause, with the evidence, not a guess
Would love feedback from anyone running multi-service or microservice backends, especially where you've been burned by "looks fine in isolation, breaks in prod" bugs.
Happy to answer anything about the architecture in the comments.
Report
Would love to see a replay mode that lets you save a captured request and rewind through the same trace later, makes it way easier to share weird repros with teammates without them needing to reproduce the bug live.
Report
Would love to see a way to replay a captured trace against a staging environment so I can verify the fix actually resolves the original failure. That's usually the part I end up doing by hand after these tools surface the issue.
Report
Finally something built for the messy distributed systems I actually work on. Watching real requests across services through a CLI feels way more natural than copying traces into a chat window.
CodeNearby
Would love to see a replay mode that lets you save a captured request and rewind through the same trace later, makes it way easier to share weird repros with teammates without them needing to reproduce the bug live.
Would love to see a way to replay a captured trace against a staging environment so I can verify the fix actually resolves the original failure. That's usually the part I end up doing by hand after these tools surface the issue.
Finally something built for the messy distributed systems I actually work on. Watching real requests across services through a CLI feels way more natural than copying traces into a chat window.