Hyrax AI - The AI architect for your entire codebase
by•
Code review tools tell you what’s wrong with a PR. Hyrax finds what should improve across your entire codebase and does the work. It builds deep context around your architecture and conventions, prioritizes issues across six engineering domains, writes each fix in context, verifies it against your tests and checks, and opens a GitHub PR for your team to review and merge.
Replies
Hey Product Hunt — I'm Tyler, Chief Architect at Hyrax.
Hyrax is a code quality tool for repos shipping code faster than anyone can review it. Two halves: it reviews every new PR, and it audits the code you already have.
For new code, connect a repo and the reviews will start automatically, nothing to configure. Hyrax will review your change against the whole repository: the definitions of every type the diff references, the sibling files that do the same job, your dependencies and migration state. That's what catches "you changed 2 of the 4 places that do this" type of bug.
No more than 7 findings at a time, so you can focus on the big items first. Every finding has to ground its premise and cite the exact lines it's talking about, then survive an internal pass that argues against it, before we post it to you.
For existing code, our full audit runs ~39 tools in six categories: Security, Correctness, Maintainability, Performance, Architecture, Operations. Findings are scored by severity and effort, so easy wins surface first.
Click "Fix", get a PR. Hyrax writes the change, runs your repo's own tests and build, reviews its own diff, then opens the pull request as a draft — and flips it to ready once your CI goes green, third-party checks included. If CI fails, Hyrax reads the failure and pushes another attempt. By default nothing will merge without you.
Your coding agents get all the context we generate, too. Hyrax will profile your repository to discover architecture, conventions, how-to guides, and serve it over MCP. Same thing we use internally.
We've got a free plan with no credit card needed. $30 credit at signup that never expires, and $10/month of credit. Full catalog, unlimited repos and teammates, MCP, everything included. Paid is $30/seat and the whole amount comes back as credit.
What's the last false positive a review bot cost you an hour on?
👉 hyrax.dev
@tyleryates I like that Hyrax checks the wider repo context instead of only looking at the changed lines. That could catch issues that are easy to miss.
I loved the idea behind it and Just wanted to know what type of programming languages and codebases it currently supports?
@jitendra_bhatt thank you for the comment! Hyrax supports all languages, feel free to run on any repo you have in GitHub. We would love feedback once you have gone through the platform
If you would like to get a more in-depth view of the app, take a look at the Hyrax demo here from @tyleryates
Website's crashing
@pulkit_gaur thank you for letting us know. Unfortunately, we do not support visitors in China due to security risks. Feel free to use a VPN to reach us.
@pulkit_gaur @tristan_benozer I am in India and still blocked?
@tristan_benozer No, I'm in India right now
@tristan_benozer I'm reaching a 403: Forbidden response from Australia?
writing the fix and opening a PR is the part that makes me most curious about scope control. For a codebase-wide sweep across six domains, is there a way to say "only touch performance issues this run" or does it always propose fixes across every domain it finds, leaving the team to triage which PRs to even look at first?
@galdayan That's a great question and it's something we've been thinking about internally. At the moment no but we may release that later this year. Are you looking to run one of the domains at a time and if so which one would be the most important to you?
@tristan_benozer performance, for us, easily. we run a voice pipeline where latency is the whole product experience, a couple hundred extra milliseconds is the difference between a call feeling natural and feeling like a bad phone line. a general "review everything" sweep would probably bury the one performance regression that actually matters under a pile of style and security nits that can wait. good to hear it's on the roadmap, that's the domain I'd want first if you ever do let people pick.
Congrats on the launch! Really like the idea of proving a bug with a failing test before reporting it. Quick question: any plans to support GitLab or Bitbucket, or is it GitHub-only for now?
@nrupen_mandava Just GitHub for now but we are looking at others in the future. Keep an eye out for the changes!