Built for engineers who are tired of repeating themselves in PR's.

Contributing guides, security policies, testing requirements — the docs exist, but they don't enforce themselves. The same comments show up on PR after PR. The same sections get pointed to over and over.

CodeReview AI reads your team's documentation and enforces it on every PR automatically. Every word, every section, every update. Findings cite the exact doc and line so the conversation starts at the right place instead of from scratch.

PRs arrive with a clear stoplight within two minutes: red means fix this, yellow means heads up, green means good to go. The obvious stuff is already flagged. Reviews can focus on architecture and judgment calls. The docs review themselves. And when a comment comes up that isn't in the docs yet — update the doc. CodeReview catches it next time automatically. That comment never needs to be made again.

How it works.

Setup takes two minutes. Install the GitHub App and CodeReview auto-detects your repo, scans the tree for documentation, classifies every file, and shows live progress file by file. Multi-topic docs get decomposed into gate-specific sections with exact line ranges. Confirm what it found and you're live.

Five gates run on every PR — Security, Architecture, Legal, Style, Onboarding — each reviewing the diff against your docs for that category. Findings are tagged 🔎 doc-backed or 💭 AI suggestion so the team knows what's policy and what's observation.

Every finding goes through a two-pass hallucination guard before it reaches the PR. If the system can't prove it came from your documentation, it gets killed. No fabricated advice. No phantom rules. Docs stay fresh automatically — someone updates a policy, CodeReview picks it up before the next review. No stale rules. No re-setup.

Results post as one PR comment. Traffic light at the top. Doc name, line range, and violation for every finding. Expandable details. Clean, scannable, actionable.

Tested against real repos — Cal.com, Stripe, Next.js, Supabase — with messy multi-topic docs, strict contribution policies, and large codebases. Currently in beta and free to use. Actively looking for teams to put it through its paces, break things, and help shape what comes next.