Introducing Free AI Code Review for Open Source Projects

Today we're making Revix AI free for open source.

Maintainers are the clearest case of the problem we built Revix AI to solve. A public repo takes contributions from people who have never read the rest of the codebase, at a pace nobody is being paid to keep up with. The code arrives faster than anyone can review it, and the review queue becomes the bottleneck on the whole project.

From today there are two ways to use Revix AI without paying anything:

• Revix free: two weeks of AI reviews across every repo you connect, public or private, with unlimited seats and nothing held back.
• Revix for open source: connect a public repository and it gets reviewed for free, every day, with no card and no clock running out.

Free review is only worth having if it reads the whole repo

Free AI review is not hard to find anymore. Most of it is diff-based: the reviewer reads the patch, checks it for bugs and style, and posts. For a typo fix, that is genuinely enough.

It is not enough for a first-time contributor. The single most common thing wrong with an outside contribution is not a bug. It's code that works fine on its own and ignores how the project already does that exact thing: the shared helper it should have called, the middleware it reimplemented inline, the convention every other file in the repo follows.

A reviewer that only reads the diff cannot catch any of that, because it never saw the files where the pattern lives. We wrote about this at length in Repo-Aware AI Code Review vs. Diff-Based Review.

Revix AI reads your repository first and derives the conventions your project already follows into rules. Then it reviews each pull request against them. For a maintainer, that is the difference between a bot that restates the diff and one that answers the question you actually have about an incoming contribution: does this belong in this codebase?

What your project gets

Connect a public repo and every pull request gets, automatically:

• A summary and a risk level the moment it opens, so you know what you're walking into before you read a line
• Comments labeled with Conventional Comments, so a blocker and a nitpick never look the same to a contributor
• Findings checked against your repo's own conventions, not a generic style guide
• Rules that improve as you correct them, so the same thing never gets explained twice

less time spent reviewing

44%

more PRs shipped per window

+31%

more issues caught before main

+58%

Those numbers come from a 30 day pilot with a production team. The shape of the win is the same on a public repo, it just lands on one or two people instead of a whole engineering org.

Contributors see the feedback before you do

The part maintainers tend to underrate: the review lands on the pull request within minutes of it opening, while the contributor is still in context.

Most of what you would have written by hand is already there when they look. They fix it, push again, and the version that reaches you is the second one. The exchanges that used to eat a week of back and forth mostly stop happening, and the ones left are the design questions that were always worth your time.

For a project that gets contributions from strangers, that is the whole difference between a PR queue that clears and one that grows.

Your code and your contributors' code

Nothing you connect is used to train models. Every review runs in isolation and nothing is kept afterward. That holds the same on the free open source plan as on a paid one.

Connect a repo

Revix AI works with GitHub and Bitbucket today, GitLab is coming. Setup is signing in and picking a repository.

Point it at your project and see what it catches on the next contribution: revix.ai