Start by deciding whether you want an AI reviewer added to your existing pull request flow or a platform where people do the reviewing. Most options are AI reviewers that comment on pull requests, and the main differences are which Git host they support, whether you can self-host, and how much of your codebase they read.
The grid comes first on this page because the split matters more than any single feature. Eight of the ten add an AI reviewer to your pull requests; two are platforms built for review by people, and they're marked as such.
AI reviewers and human review platforms, side by side
| Tool | Approach | Good for |
|---|---|---|
| cubic | AI code review | Complex codebases |
| Entelligence | AI review and engineering intelligence | Leads who want review data and engineering insight |
| Gerrit Code Review | Open source review (human-led) | Git teams that want review by people |
| Gitar | Review and CI fixing agent | Teams whose builds fail often |
| Graphite | Code review platform | Teams whose code lives on GitHub |
| Kodus | AI code review | Teams that prefer open source |
| Korbit | AI reviewer | GitHub and Bitbucket teams |
| Macroscope | AI code review | Leads who want status updates too |
| Propel | AI reviewer | Teams wanting a tech-lead style reviewer |
| Review Board | Code and document review (human-led) | Teams reviewing documents as well as code |
A note on the first column: anything marked human-led manages how your engineers review each other's code. It won't write review comments for you, and that's the point of including it.
The ten review tools, alphabetically
Sorted by name and in no particular order of preference. Each line repeats a claim the vendor makes about itself, so treat them as starting points for a trial.
cubic
Best for: Catching hard-to-find bugs, which the vendor says it finds in pull requests and across your entire codebase.
Entelligence
Best for: Teams after what the vendor calls one engine and one memory for AI code review and model routing.
Gerrit Code Review
Best for: Patchset-based review; the project calls itself the open source, patchset based code review platform for Git-based development.
Gitar
Best for: Diagnosing CI failures and applying fixes on its own, which the vendor pairs with what it calls high-signal AI code review.
Graphite
Best for: Shipping what the vendor calls higher quality software, faster.
Kodus
Best for: Self-hosting for free, or the cloud with your own model keys at what the vendor calls zero markup.
Korbit
Best for: Faster review cycles; the vendor says it reviews your pull requests first, so fewer bugs get through.
Macroscope
Best for: Checking every change for correctness, security, tests and regressions before it lands, according to the vendor.
Propel
Best for: Reviews that come with fixes and guidance, as the vendor says it reviews, fixes and guides your development team.
Review Board
Best for: Review at scale while staying on-prem and private, per the project's own description.
Who still sells code review as the main product?
Several names that come up when you search for AI code review no longer lead with it. On their own sites, Ellipsis and Bito lead with managed cloud agents and a model router, so they aren't rows here. Gitar is still on the list, but it now sits on Sonar's site and is presented as a Sonar product, so if you buy it, you're buying into that wider product family.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
The two human-led platforms are here as the older route. Review Board describes itself as a review tool for teams, available on-premises or hosted. Gerrit Code Review says it's a web-based code review system for Git repositories. If your problem is that reviews are slow or inconsistent rather than that reviewers miss bugs, a better human process may do more than another bot.
Reading the AI reviewers by approach
Context beyond the diff. Propel says it reads pull requests with an understanding of your architecture, codebase and policies. cubic, the vendor says, checks pull requests against your team's own rules. Entelligence, the vendor says, checks pull requests against past incidents.
Extras around the review. Korbit, the vendor says, gives context-based reviews on pull requests. Graphite says it has AI review built into its platform. Macroscope adds automated pull request descriptions to its reviews, according to its site.
For a wider look at the tooling around review, my guide to developer productivity tools covers what to look for, and my Claude Code alternatives list covers agents that can also read a diff before you open a pull request.
Build an AI reviewer or review platform that fits here? Ask for it to be considered the next time this page is reviewed.
Practical advice for a fortnight's trial
Don't choose from a demo repository. Install two reviewers on a live repository for the same fortnight and let both comment on the same pull requests, so you're comparing them on your own code. Make sure the pair you pick both support your Git host, and if your code can't leave your network, start with the tools that say you can self-host.
Track three things while they run: how many comments your engineers act on, how many they dismiss as noise, and whether either tool caught something a person missed. A reviewer that leaves fewer comments that people act on is worth more than one that leaves dozens nobody reads.
- Pick a repository with steady traffic and a mix of small and large pull requests.
- Agree which comments count as useful before the trial starts, so the result isn't a matter of mood.
- Turn both tools off at the end and ask the team which one they missed.
Want your tool on this list? Featured spots and reviews here are paid partnerships. Here's how to partner with me.
Frequently asked questions
Can an AI reviewer replace human code review?
No. Treat it as a first pass that catches routine problems so people can focus on design, intent and anything risky. Keep a human approval step on every change.
Do AI code reviewers work with Bitbucket and GitLab?
Support varies by tool, so check each vendor's own integrations list for your Git host before you start a trial.
Is my code used to train the reviewer's model?
Check each vendor's security and data pages. Self-hosted options keep code on your own servers, which some teams need.
What's patchset review?
It's a workflow where each change is uploaded as a patchset that reviewers comment on and approve, and new versions of the change are added as further patchsets until it's merged.
How do I stop an AI reviewer from flooding pull requests with comments?
Check during the trial whether you can set rules, ignore paths and choose how strict it gets. Start on a single repository and widen it once your team trusts the comments.
