AI agents for code review
Review pull requests for bugs, security and style before a person merges.
10 agents, each with its inputs, outputs and review level checked by a person.
Code review agents read pull requests and leave comments on bugs, security issues, style and missing tests before a person merges. They connect to your code host and either review automatically on every pull request and push, or review only when someone asks, by mentioning the agent, adding a label or running a command. Automatic review gives consistent coverage but more comments to read; review on request keeps the noise down for teams that only want it on larger changes. Many products support both.
Use them to catch routine issues so human review can focus on design and intent. The listings record the hosts each agent connects to, how it is triggered, what it comments on, whether it can suggest or push fixes, how rules are set and whether a person accepts each comment. None of the agents listed here merge on their own.
Check whether the agent can be tuned with your own rules and conventions, whether it summarises the change as well as commenting, and how it is priced.
The agents
- Claude Code Reviewby AnthropicReviews GitHub pull requests with several agents and posts verified findings ranked by severity.Pricing not publishedNeeds configurationHuman reviewDetails
- Cursor Rolloutsby CursorWrites a monitoring plan before merge and flags regressions after deploy, with a revert pull request for approval.Pricing not publishedNeeds configurationOptional reviewDetails
- Cursor Security Reviewerby CursorChecks every pull request for security vulnerabilities and proposes a fix for each finding.Pricing not publishedNeeds configurationHuman reviewDetails
- IBM Bobby IBMCoding agent for the IDE and command line, with subagents and add-on packages for Java and IBM i modernisation.Subscription, from $20/moDeveloper setupOptional reviewDetails
- Augment Codeby Augment CodeCoding agents grounded in a context engine for reviews, tests, migrations and ticket-to-PR work.Subscription, from $20/moNeeds configurationOptional reviewDetails
- CodeRabbitby CodeRabbitReviews every pull request with comments, one-click fixes and pre-merge checks.Subscription, from $24 per developer/mo billed annuallyNeeds configurationHuman reviewDetails
- Cursorby CursorCoding agent in an editor, terminal and the cloud that plans, builds and tests changes for review.Freemium, from $8/moWorks instantlyOptional reviewDetails
- Factoryby FactoryDroids that take a backlog item, implement it, review it and ship it through your existing tools.Subscription, from $20/moDeveloper setupOptional reviewDetails
- GitHub Copilot coding agentby GitHubAssign a GitHub issue to Copilot and get back a pull request to review.Freemium, from $10/moWorks instantlyHuman reviewDetails
- Qodoby QodoAgentic pull request review across repositories, in the IDE, CLI and Git workflow.Subscription, from $30/moNeeds configurationHuman reviewDetails
No agent matches. Clear the search or choose another filter.
Related
Groups that share agents with this one.
Questions
Does it review every pull request automatically or only when asked?
Most can do either. Automatic review runs when a pull request opens and again on new pushes. Review on request runs when someone mentions the agent, adds a label or runs a command. Many teams start automatic on one repository and adjust from there. The listing records the triggers the vendor documents.
What can an AI code reviewer catch, and what does it miss?
They are good at local problems in the changed code: likely bugs, missing checks, unsafe patterns, inconsistent naming and missing tests. They are weaker on whether the change is the right design, on business rules nobody wrote down and on effects in parts of the system the diff does not touch. Treat comments as suggestions and keep a person responsible for the merge.
Can it push fixes to the pull request?
Some offer suggested changes you can accept with a click, and some can commit a fix to the branch or open a follow-up pull request when asked. Others only comment. The outputs and review level on each listing show what the vendor documents.
How is AI code review priced?
Usually per seat, counting developers who open pull requests, or per repository, sometimes with a free tier for public or small projects and usage limits on larger plans. Each listing shows the published model with the date it was checked.
Will a review agent block merges?
Only if you configure it as a required check. By default the agents here comment and a person decides; the review level on each listing reflects that.
Does it see my whole codebase?
It reads the repository it is connected to in order to review changes in context. The vendor's security page, linked from the listing, describes retention; the catalogue does not verify it.