Autonomous code review and fixing for engineering teams. Hyrax connects to your GitHub, audits the whole codebase across security, correctness, performance, maintainability, architecture, and operations, writes the fix, and sends it as a pull request you review and merge. Before it reports a problem it writes a failing test to prove the problem is real. Before a fix reaches you it runs the repo's own tests, build, and lint. The free plan is a mini audit, 5 findings, and 5 fixes a month.

Hi everyone, I'm Tyler, I built Hyrax. We leaned into AI coding tools and our output went up, and so did the amount of broken code that came with it. Reviewing and fixing all of it by hand didn't keep up, so the bottleneck moved from writing code to checking it. It began as a handful of scripts and skills we ran internally to review our own PRs before they merged. That grew into a CLI, then the product we use every day and are now sharing. The part that changed most as we built it was trust. Telling a model to review and fix something gives you confident changes you can't verify, so the whole design became about proof. Before Hyrax reports a finding it writes a test that fails on the bug. Before a fix reaches you it runs your own tests, build, and lint, and a second pass reviews it. Everything comes back as a pull request, so it never pushes to main and nothing merges without you. The free plan is a mini audit of your repo, 5 findings, and 5 fixes a month. GitHub only right now. Would love your feedback, especially on false positives.

One thing that would help me trust it more: a quick severity score on each finding so I can triage what to look at first instead of reading every explanation. Even just a low/medium/high tag tied to the category would make the PR way faster to review.

Tristan Benozer

Continuous code remediation for Engineers

Continuous code remediation for Engineers