AI bots broke the review page
Tools like Copilot, CodeRabbit, and Greptile are genuinely useful — they catch real issues. But they pile on 10, 20, 50 comments per push, and GitHub's UI was never built for that volume. The PR page grinds to a crawl, threads collapse into noise, and the actual code review — the human part — gets buried.
And this is only the beginning. AI bot reviews will keep getting more powerful — at a pace we've never seen before. If reviewing a 50-comment bot thread is painful today, imagine 200 comments from three different bots tomorrow.
So we built a native app
GitHub will probably fix this eventually. But the web UI was never designed for this volume. Rather than waiting, Hayato built a proper tool — not a web app, not an Electron wrapper — a native SwiftUI application built for raw performance. Keyboard-driven. Fast. Cuts through the noise.
We build the GUI, you bring the AI
PR never collects, intercepts, or routes your code through our servers. The app calls whichever AI CLI you already have installed — Claude, Gemini, or Codex. Your credentials, your tokens, your rules.
Apple Intelligence changes the equation
Claude, Gemini, and Codex consume your own API tokens. Apple Intelligence runs entirely on-device at zero token cost. It won't give you the sharpest results, but it's often enough to summarize a PR or classify changed files — the kind of lightweight triage that saves time without burning your budget.
Worried about tokens?
Use Apple Intelligence. On-device, zero cost, good enough for summaries and classification.
Want the sharpest AI?
Pick Claude, Gemini, or Codex. Best-in-class results, powered by your own API credentials.
Switch providers anytime. No lock-in, no migration.