AI code review for your pull requests
A second pair of (AI) eyes on every pull request. This matters most when most of your code is written by an agent.
Not hands-on tested yet. The picks below are based on vendor documentation and pricing we verified on 19 Oct 2026. Test results will appear here once we've run our standard task. We don't publish made-up test results.
CodeRabbit
Dedicated PR reviewer covering GitHub/GitLab PRs, the CLI and the IDE, with one-click fixes, pre-merge checks and slop detection. Free forever on public repositories.
source, checked 19 Oct 2026
Greptile (Starter)
A free tier for a single developer, with unlimited repos and 50 review credits a month (a base review costs 1 credit). Enough for a solo builder's private projects.
source, checked 19 Oct 2026
Turning on org-wide Copilot code review for unlicensed users without a budget
GitHub can review PRs from users without a Copilot licence, but that usage is billed to the organisation as AI credits, and since 1 June 2026 review workflows also use Actions minutes. Set budgets first.
The full stack
- Write: Your AI coding agent or builder
Generate the change in a branch, never directly on main. - Review: CodeRabbit (or Greptile)
Automatic review on every PR: bugs, security and style. - Gate: Pre-merge checks + CI tests
Block merges until review findings and tests pass.
Also considered
| Tool | Why it's not the pick | Pricing (verified) |
|---|---|---|
| GitHub Copilot code review | Convenient if you already pay for Copilot; PR review is part of the plans and consumes AI credits. | Free (limited); Pro $10/mo; Pro+ $39/mo; Max $100/mo. GitHub lists total monthly AI credits of $15 (Pro), $70 (Pro+) and $200 (Max), including a variable flex allotment. (source, 19 Oct 2026) |
| Cursor Bugbot | If you already pay for Cursor Premium or above, AI code reviews via Bugbot are included in the plan, so try it before adding another tool. | Included from Premium $30/mo (not on Free or Plus). Third-party models billed separately at API list price. (source, 19 Oct 2026) |
Test notes
Pending hands-on test
Pending hands-on test. Planned standard task: open the same PR, containing 5 seeded bugs (an injection, a missing auth check, an off-by-one, a leaked secret, a race condition), against each reviewer and count true positives, false positives and the time to the first comment.
Why this matters for vibe coders
AI agents write plausible code quickly, and that includes plausible bugs. An automated reviewer on every PR is the cheapest insurance you can buy.
How to choose
- Public or open-source repo: CodeRabbit is free.
- Solo, private repos, £0 budget: Greptile Starter.
- Small team that ships daily: CodeRabbit Essentials or Team.