CodeRabbit

AI code review per PR — free forever for OSS/public repos

CodeRabbit

Role

LLM-driven code review on every pull request across the family. CodeRabbit reads the diff, understands intent, and posts inline comments calling out:

• Logic errors (off-by-one, wrong condition, missing null check)
• Security smells (unsafe input handling, injection vectors, leaked secrets)
• Design issues (tight coupling, missing error paths, race conditions)
• Style drift from the rest of the file

It catches the class of bugs Dependabot and biome both miss — the ones that need a model to understand the code, not just match patterns.

Free tier

Free forever for public repos:

• 200 files/hour throughput
• 3 reviews/hour
• 50 conversations/hour
• Unlimited repos
• All review features (chat, walkthroughs, summaries)

The family is entirely OSS, so this tier covers everything.

Card / subscription required?

NO for OSS use.

Alternatives

• Greptile — similar AI review tool, also has an OSS-friendly tier
• Sourcery — Python-leaning, less helpful for the TypeScript-heavy family
• Pull-request-reviewer GitHub Actions wrapping ChatGPT — DIY route; we’d own the prompt engineering

Swap cost

Low. Both alternatives are GitHub-app installs that comment on PRs; replacing is a one-time install + uninstall in the GitHub org settings.

Why this is our pick

CodeRabbit’s free OSS tier is unbounded in repo count, which matches the family’s repo-per-site shape. The reviews are noticeably stronger on TypeScript than Sourcery, and the conversational follow-up (“why?”, “how do I fix?”) gives a useful learning loop on top of the review itself.

Cross-refs

• services/code/code-quality/dependabot
• services/code/code-quality/sonarcloud
• rules/match-surrounding-style
• runbooks/clean-install
• decisions/code-quality-stack

Edit on GitHub · Back to index