We just launched Proton, a GitHub app that listens to PR review comments and suggests code changes to address them. It creates a new PR based on top of your working branch, so you can simply review and merge.

How is it different from other AI code review tools? Others focus on pointing out issues, whereas Proton focuses on addressing them. Although some of them can also suggest fixes, they tend to only work on nearby lines or within a single file. But real-world feedback often involves cross-file changes like “Let’s extract this to a separate component”, or “We should follow the same pattern of doing things in file X, Y and Z”. Proton has full repo context, so it can handle these kinds of feedback.

Want to see it in action? Here’s a short demo: https://youtu.be/zDEfw-R2jWc, and there’s the PR shown in the demo video: https://github.com/proton-codes/demos/pull/2

It’s free, takes two clicks to install, and works out of the box. Here’s the install link: https://github.com/apps/proton-app

We'd love to get some early users and hear your thoughts — reply here or email us at [support@proton.codes](mailto:support@proton.codes)

Cheers!