CodeRabbit reviews pull requests on GitHub, GitLab, Azure DevOps and Bitbucket, and checks local changes from VS Code-compatible editors or its CLI. Paid plans are billed per developer who opens pull requests, and public open-source repositories get Team features free.
Used this tool? Rate it
Used this tool? Rate it
CodeRabbit is an AI code review tool that comments on pull requests inside your Git platform and can also check local changes from an editor or a terminal. The company describes it as an AI-first pull request reviewer with context-aware feedback, line-by-line code suggestions, and real-time chat. It does not replace a coding assistant; it works at the review step, reading a change, flagging likely bugs, proposing fixes, and answering follow-up questions where the discussion already happens.
Higher plans add queue prioritization (Triage), security review and finishing actions, and a separately billed CodeRabbit Agent handles cloud coding tasks and Slack. As captured on September 28, 2026, the CodeRabbit homepage lists 17K customers and 6M repositories.
Harjot Gill started CodeRabbit in early 2023 as an AI-powered code review platform. In September 2025, independent reporting covered a $60 million Series B that valued the company at $550 million.
A review is more than a diff read in isolation. CodeRabbit builds a codegraph, a deterministic map of every node a change can affect. It can also pull in context from PRDs, issues and internal docs relevant to the work, so a comment can weigh a change against the task it belongs to. For the review itself, CodeRabbit combines multiple models on difficult problems instead of relying on a single one.
Repository-level settings live in a .coderabbit.yaml file, which must be located in the root of the repository. To avoid spending reviews on noise, automatic review controls can pause incremental reviews after 1 or 2 reviewed commits, require a ready-for-review label, exclude titles such as WIP, or leave every review to manual requests. These controls matter because each counted review draws on an hourly allowance, as described under Limitations.
The CLI option --uncommitted limits the review to staged and tracked local edits, and files not added to Git are skipped unless --include-untracked is passed. This makes it possible to clear obvious problems locally before a pull request exists.
Good fit
Less suited
CodeRabbit pricing is per developer per month. CodeRabbit offers Free and OSS access plus four paid plans: Essentials, Team, Advanced and Enterprise; the pricing page itself shows only the four paid plans and mentions open-source access in its FAQ.
| Plan | Billed annually | Month-to-month | Main additions |
|---|---|---|---|
| Free | No charge | No charge | PR summaries; reviews through the VS Code extension and CLI |
| Open source | No charge | No charge | Team features on public repositories |
| Essentials | $24 | $30 | AI reviews on PRs and CLI, one-click fixes, learnings, MCP connections, Jira and Linear integrations |
| Team | $48 | $60 | Triage, custom pre-merge checks, finishing touches, post-merge actions |
| Advanced | $72 | $90 | Security review on every PR, continuous security monitoring |
| Enterprise | Custom | Custom | Self-hosting, SSO, custom RBAC, audit logging, API access |
Annual billing saves 20% on the self-serve plans compared with month-to-month billing. Compared with Essentials, Team adds Triage, expanded Change Stack capabilities, custom pre-merge checks, unit-test generation, merge-conflict resolution, post-merge actions and higher connection and multi-repository limits.
New organizations start with a 14-day Advanced trial that includes 10 PR, IDE and CLI reviews per developer per hour plus 50 chat messages per hour. Organizations that do not subscribe after the trial revert to Free with lower rate limits. Currency also narrows the choice: USD-billed organizations can choose Essentials, Team or Advanced, while INR-billed organizations can choose Essentials or Team.
Several charges sit outside the seat price:
Plan upgrades take effect immediately with prorated billing, while downgrades take effect at the end of the billing cycle. The terms of service state that expiration or termination does not entitle the customer to a refund, except termination for cause by the customer.
Independent reporting in 2025 named Graphite and Greptile as startup rivals of CodeRabbit. The same coverage noted that AI coding assistants such as Claude Code and Cursor also offer AI-powered code review.
Graphite describes itself as helping teams on GitHub deliver higher-quality software. Its feature list covers stacked PRs, a PR page, AI code review, chat and a merge queue. Graphite is free for the first 30 days with no credit card required. It suits teams that want review and merging workflow in one GitHub-centred tool, whereas CodeRabbit also covers GitLab, Azure DevOps and Bitbucket. Greptile starts free with 50 monthly credits and offers team and self-hosted options. Review built into Claude Code or Cursor stays inside the tool that produced the change, whereas CodeRabbit runs as a separate reviewer on the pull request.
In 2025, independent security researchers reported remote code execution on CodeRabbit production servers, leaked API tokens and secrets, and read and write access to 1 million code repositories, including private ones. The researchers state that the issues were quickly remediated in January 2025.
Yes, for public repositories: open-source projects get Team features without a paid subscription. The allowance varies with the project's popularity, and the privacy policy carves open-source code out of its no-training statement.
The documentation and the terms say no, including for OpenAI and Anthropic, which receive code for reviewing. Cached archives, embeddings and learnings can still be stored unless retention is turned off.
GitHub, GitLab, Azure DevOps, Bitbucket Cloud and Bitbucket Data Center for pull requests; VS Code, Cursor and Windsurf for the extension; and a terminal CLI.
No. Only developers who create pull requests are charged, and usage-based reviews, Agent minutes and full Security Scans are billed on top when enabled.
New reviews wait until earlier ones age out of the rolling window, unless usage-based reviews are enabled, which continue eligible reviews at $0.25 per reviewed file up to the admin's monthly cap.
