CodeRabbit Review

CodeRabbit sits on the pull request, not in your editor. You open a PR, it posts a summary, then it leaves inline comments with suggested patches you can commit from the review. That is a different job from Copilot-in-the-IDE. The question is whether those comments catch real bugs without burying the human reviewer in nits.

This review is for engineering leads and developers deciding if a dedicated reviewer is worth another seat on top of their coding assistant. Pricing below is from coderabbit.ai/pricing as of September 2026. I have a partner link at https://coderabbit.link/louis-dupple-com.

There is no extra percent off on that link, just the referral. I am not claiming a bake-off on your repos. I am laying out the published plans, the usage meters that actually change the bill, and how Greptile, GitHub Copilot, Graphite, Qodo, and Bito differ on paper.

One billing habit to get straight before the seats look cheap. CodeRabbit sells a per-developer month, then meters bursts ($0.25 a file) and agent minutes ($0.40). A ten-person squad on Essentials annual is $240 a month before anyone hits the hourly cap.

That is the number to put next to Copilot Business, Greptile Pro, and a Graphite Team seat, not the $24 sticker alone.

CodeRabbit is an AI-first pull request reviewer built by CodeRabbit (founded 2023, San Francisco, with a large team in India). You install it on GitHub or GitLab. It reads the diff, writes a summary, and comments line by line.

Under the LLM it also runs linters and SAST tools, so a single check can cover style, known vulnerability patterns, and the obvious logic slip.

The 2026 plan names matter because older posts still say Pro and Pro Plus. Essentials is the old Pro. Team is the old Pro Plus. Existing subscriptions were renamed in place.

The product around those seats now includes agentic reviews on PRs and the CLI, one-click fixes, a learnings store so it stops repeating nits you dismissed, loops with coding agents, MCP connections, and chat. Team adds triage, custom pre-merge checks, finishing touches (unit tests, merge-conflict help, simplify), and post-merge actions.

Advanced adds blast-radius analysis, architectural impact, and a security review on every PR plus continuous security monitoring.

Public open-source repos stay free. That is a real plan, not a teaser: install on a public repo and reviews keep running. Private repos and the deeper features are what you pay for.

Git host coverage is GitHub and GitLab on the public pages. Bitbucket teams should look at Bito instead of hoping a silent adapter appears. Seats follow people who open pull requests, which is friendlier than billing every GitHub user in the org, and you can reassign a seat when someone leaves. That is the procurement detail finance will ask for after the trial.

You connect the Git host, pick repos, and CodeRabbit reviews new pull requests. On each PR it posts a summary, then inline comments. Suggested fixes can be applied in one click. Learnings let you teach it "stop flagging this" so the next PR is quieter.

You can also run reviews from the CLI and keep an agent loop going while you work in Cursor or Copilot.

The meter people miss is the hourly review cap. There is no hard monthly ceiling on how many PRs you can open. Each plan allows a burst of reviews per developer per hour: 5 on Essentials, 8 on Team, 10 on Advanced, 12 on Enterprise, subject to a fair-use policy.

If a squad merges a huge stack before lunch, you hit the cap. Admins can then turn on usage-based reviews (automatic, on demand, or off) with a monthly spend cap. Continued reviews bill $0.25 per reviewed file.

The other meter is CodeRabbit Agent at $0.40 per agent minute for cloud coding tasks, Slack, and automations, after any trial or free minutes. Enabling paid Agent usage opens it for the org, again with a spend cap.

Seats are billed for developers who create pull requests, and you can reassign seats. You are not paying for every GitHub account that never opens a PR.

Noise is the other operational problem. A reviewer that restates the diff trains humans to collapse the thread. Learnings exist to stop that, but only if someone on the team actually dismisses junk and lets the store update.

Custom pre-merge checks on Team are how you encode house rules (no new raw SQL in the API package) instead of arguing with a generic model. If nobody will own that config, you will get a chatty bot and blame the vendor.

AI code review is sold as a seat per developer, then a second meter for bursts: extra files, credits, or lines of code. CodeRabbit, Greptile, GitHub Copilot, Graphite, Qodo, and Bito each publish an entry price.

Enterprise is a custom quote at CodeRabbit, Greptile, and Bito, and GitHub Copilot prices Business and Enterprise seats separately from its individual plans. The cheapest published dedicated seat is Bito Team at $12 per seat per month, billed annually.

CodeRabbit Essentials is $24 per developer per month billed annually, or $30 month to month, and usage past the hourly cap is $0.25 per file, and the Agent is $0.40 per minute.

The other jumps are Greptile at about $1 per extra credit, Bito at $5 per extra 1,000 lines after about 5,000 per seat, Graphite Team at about $40 per user per month billed annually for unlimited reviews, and Qodo Pro Team from about $30 per month for up to 30 users.

CodeRabbit is a strong fit for teams that want a drop-in reviewer on every private PR and already live on GitHub or GitLab. Open-source maintainers should install it on public repos at $0 and stop there unless they also have private code.

Essentials at $24 annual is the default paid seat for a squad that wants comments and one-click fixes. Move to Team at $48 when you need triage and custom checks. Advanced at $72 is for teams that will actually use blast-radius and the security pass, not for the logo.

It is a weaker fit if you already pay for Copilot and only wanted shallow PR comments, because Copilot review is then "free" in the sense that you are not adding a second vendor.

It is also a weaker fit if your whole process is stacked diffs (Graphite owns that workflow) or if you care more about whole-repo graph context than a tunable comment stream (Greptile's bet). Watch the $0.25/file overage and the $0.40/min Agent if a few people run huge refactors or long agent sessions. Those meters are how a predictable seat becomes a surprise invoice.

Open-source maintainers should install on public repos and stop. That is a complete product at $0. Mixed orgs (public plus private) can keep OSS free and pay only for the private seats. Monorepos with 40-file refactors are exactly where the hourly cap and $0.25/file toggle matter; decide the admin policy before the first merge week, not during it.