On this page
The best PR review bots for GitHub in 2026 are CodeRabbit and Greptile for hosted AI review, GitHub Copilot code review if you already pay for Copilot, and the repowise bot or Reviewdog for repeatable, rule-based comments. A PR review bot is a GitHub App or CI job that comments on pull requests automatically. The right one depends on how much noise you will tolerate, whether the bot needs to see beyond the diff, and what each seat or review costs at your volume.
The short list, with prices checked on 10 November 2026:
- repowise PR bot: deterministic comments built from the code graph and git history, free on every repository, quiet when there is nothing to flag, and optional AI review at about 15 cents a pull request on average.
- CodeRabbit: the broadest general-purpose AI reviewer, free on public repositories.
- Greptile: AI review against an index of the whole codebase, with self-hosting on Enterprise.
- GitHub Copilot code review: no new vendor for teams already on a paid Copilot plan.
- Cursor Bugbot: review that hands its findings to Cursor's agents to fix.
- Claude Code Review: Anthropic's multi-agent reviewer, the deepest and most expensive per pull request.
- Codacy: a code quality platform with an AI reviewer included.
- Reviewdog: open source, turns your linters' findings into PR comments.
We build repowise, so weigh our entry accordingly. The measurements we publish about it, with their methods, are on the benchmarks page.
How we evaluated PR review bots
We compared each bot on four things a team decides on, using each vendor's own pricing page and docs, fetched on 10 November 2026:
- Comment discipline: does it stay quiet when the diff is clean, and update its own comments instead of piling up new ones?
- Context: does it see only the patch, or also callers, ownership and history?
- Repeatability: does the same diff get the same comment, so a finding can gate a merge?
- Price at your size: per seat, per credit or per review, and what is free for open source.
The test that matters after that is replaying about 20 of your own merged pull requests through the shortlist and keeping the bot whose comments you would have acted on. Installing as a GitHub App also helps with access, because apps get scoped repository permissions.
Comparison table
| Bot | Comment style | Price (November 2026) | Free on public / OSS repos | Self-host |
|---|---|---|---|---|
| repowise PR bot | Deterministic Signals by default, optional AI review | Signals free on every repo, no model call; AI review about 15 cents a PR on average from credits, no per-seat or per-repo fee | Yes, public repos | Bot is hosted; engine and CI gates are open source |
| CodeRabbit | LLM review, incremental per push | $24 / $48 / $72 per developer per month annual (Essentials / Team / Advanced) | Yes, public repos | Enterprise, 500+ seats |
| Greptile | LLM review with codebase context | Free for 1 developer (50 credits a month); Pro $1 per credit, no seat fee | Qualified non-commercial OSS | Enterprise plan |
| GitHub Copilot code review | LLM review inside GitHub | Paid Copilot plan; about $1.50 of AI credits per Balanced review, plus Actions minutes | No | No |
| Cursor Bugbot | LLM review with autofix | $1.00 to $1.50 average per run, on Cursor Premium ($30 a month) or Teams ($30+ a user) | No | No |
| Claude Code Review | Multi-agent LLM with verification | Claude Team or Enterprise plan; about $15 to $25 a review (launch estimate) | No | No; runs on Anthropic infrastructure |
| Codacy | Static analysis plus AI reviewer | Team $18 per developer per month (yearly) | Yes, open-source projects | Ask the vendor |
| Reviewdog | Your linters' findings | Free, MIT licensed | Yes | Yes, runs in your CI |
Scroll the table sideways to see every column.
For yearly bills by team size, see AI code review pricing compared or the cost calculator.
What makes a PR bot worth keeping
A useful bot stays quiet when the diff is clean, comments on the changed lines, does not repeat itself across pushes and manages its own threads. A bot that comments on everything trains reviewers to ignore it.
Deterministic bots map rules, linters or analysis results to comments, so the same diff gets the same output. LLM bots read more code and explain a problem in plain English, but their output varies run to run. Many teams pair one of each. For the trade-off in detail, see deterministic vs LLM code review.
1. repowise PR bot
The repowise PR bot is a hosted GitHub App that builds each comment from the repository's index, with no model call in its default mode. Because the comment draws on the code graph, ownership, co-change history and file health, it says things a diff-only reader cannot: a changed function has callers outside the PR, a file is getting worse because of the complex method this PR adds, or files that usually change together are missing from the PR.

In Signals mode the same diff always gets the same comment, one comment per pull request is kept up to date on every push, the worst findings are marked on diff lines, and a pull request with nothing to flag gets no comment. A "repowise review" check on every push shows the reviewed commit and what is still open, and the health check can be required in branch protection. The "View on repowise" button opens a public per-PR page with no sign-in on public repositories; repowise-dev/repowise#4506 is a live example. The silent-by-default write-up covers the design.
AI review adds a model pass on every pull request that changes code, a check on risky claims before they post, and a deep review once when change risk calls for it. Comment /repowise recheck to re-run or /repowise review deep to force the deep pass. The engine is open source under AGPL-3.0, and its coverage, doc-drift, security and change-risk gates run as a GitHub Action in your own CI. The same index powers docs, graphs and ten MCP tools; see it on real repositories in the live examples.
Pricing: the lowest cost per pull request among the hosted bots here. Signals is free on every repository and makes no model call. AI review is paid per pull request from credits, with no per-seat or per-repo fee: about 15 cents a pull request on average, so 100 reviewed pull requests cost about $15, capped at 60 cents a PR and $5 a repository a month by default. Private repositories need a paid plan, and Teams adds a shared workspace and a pooled credit balance (repowise pricing).
Pick it if you want consistent comments with repository context, a merge gate you can trust, and AI review without a seat fee.
2. CodeRabbit
CodeRabbit is the most polished general-purpose AI review bot in this group. Reviews are free on public repositories. Per its pricing page on 10 November 2026, paid plans are Essentials at $24 per developer per month, Team at $48 and Advanced at $72 on annual billing ($30, $60 and $90 monthly), with Enterprise on custom pricing. Higher tiers add custom pre-merge checks, generated tests and, on Advanced, Blast Radius and architecture impact review.
It re-reviews after each new push and focuses on commits added since the last review, so feedback arrives in small increments (docs). It supports GitHub, GitHub Enterprise Cloud and GitHub Enterprise Server, and its self-hosted option is for Enterprise customers with 500 or more seats.
Pick it if you want the strongest hosted all-rounder with broad language support and onboarding that needs no sales call. Pricing detail: CodeRabbit pricing; comparison: repowise vs CodeRabbit.
3. Greptile
Greptile is the strongest codebase-aware AI reviewer here. It reviews pull requests on GitHub, GitLab, Bitbucket and Gitea, per its docs as of November 2026, with context from an index of the whole codebase. Pricing is credit-based: a free Starter plan for one active developer with 50 credits a month, Pro with no seat fee and prepaid credits at $1 each shared across the organization, and Enterprise with self-hosting. A Base review costs 1 credit, Plus 3 and Apex 10, and an Auto setting picks the tier per PR. Qualified non-commercial projects with an OSI-approved license are free (pricing, 10 November 2026).
Enterprise adds self-hosting, which matters when source code cannot leave your network (self-host docs), plus GitHub Enterprise Server, GitLab Self-Managed, Bitbucket Data Center and Perforce support. Credit pricing means a month of deep reviews costs more than a quiet one, so model it against your volume.
Pick it if you want AI review with whole-codebase context, or need it self-hosted. Pricing detail: Greptile pricing; comparison: repowise vs Greptile.
4. GitHub Copilot code review
GitHub Copilot code review is the PR bot you already have if your team pays for Copilot. You request Copilot as a reviewer, or set a rule so it reviews every pull request. It is included in Copilot Pro, Pro+, Max, Business and Enterprise; Copilot Free only reviews a selection in VS Code (docs).
Each review spends AI credits, which GitHub estimates at about $0.25 on average at Lite, $1.50 at Balanced (the default) and $10 at Max, in public preview, plus GitHub Actions minutes for the agentic context gathering. Automatic reviews bill the pull request author by default. The bill scales with how many pull requests you open and how large they are.
Pick it if you want to try automated review without adding a vendor. Watch credit spend for a month before enabling it on every pull request; our Copilot code review cost post works one through.
5. Cursor Bugbot
Cursor Bugbot reviews pull requests and can hand each finding to a Cursor agent to fix. It runs on GitHub, GitLab, Bitbucket and Azure DevOps, including GitHub Enterprise Server, GitLab Self-Hosted and Bitbucket Data Center, and a cursor review or bugbot run comment triggers it on demand. Team rules live in .cursor/config/bugbot.yaml, and autofix makes at most three attempts per PR (docs).
Bugbot is usage-based: Cursor's help center puts the average run at $1.00 to $1.50, with Low, Default, High and Smart effort levels. It is included from Cursor Premium ($30 a month) and Teams ($30 or more a user a month), per cursor.com/pricing on 10 November 2026.
Pick it if your team already writes code in Cursor and wants review findings to flow into the same agents. More in Cursor Bugbot pricing and alternatives.
6. Claude Code Review
Claude Code Review is Anthropic's managed, multi-agent pull request reviewer for GitHub. Several agents read the diff alongside the surrounding code, a verification step drops false positives, and findings post inline ranked by severity, with a check run that never concludes as a failure (docs). It is for Claude Team and Enterprise plans, supports GitHub Enterprise Server, and offers Quick, Standard and Deep review levels; @claude review on a PR starts one by hand.
Anthropic's March 2026 launch estimate is $15 to $25 a review on average, billed on token usage separately from the plan, and a repository added from the admin page reviews after every push by default (cost docs). It is not available to organizations with Zero Data Retention.
Pick it if a missed defect is expensive enough that a review in the tens of dollars is cheap, and you point it at the riskiest pull requests. More in Claude Code Review pricing.
7. Codacy
Codacy is a code quality platform that includes an AI reviewer. Its AI Reviewer combines static analysis with context-aware reasoning to comment on pull requests about security fixes, untested critical functions, complexity, duplication and whether the code matches the PR description. It is included in the Team plan at $18 per developer per month billed yearly ($21 monthly), alongside merge gates, and open-source projects are free (pricing, 10 November 2026). In March 2026 Codacy deprecated its AI-enhanced comments beta in favor of the AI Reviewer.
Pick it if you want quality gates, security scans and PR feedback from one vendor. See repowise vs Codacy.
8. Reviewdog
Reviewdog is an MIT-licensed tool that posts your linters' and analyzers' output as PR comments, with no AI model involved. It reports as github-pr-review comments with code suggestions, GitHub Checks, annotations, GitLab merge request discussions, Gerrit, Bitbucket code reports and Gitea reviews (repository). It runs in your own CI and costs nothing.
It knows what your analyzers know, so an architectural or cross-file problem needs another layer.
Pick it if your team already trusts its linters and wants their findings in the pull request.
How to choose a PR review bot
Start from what your team already pays for and what must stay inside your network, then price the rest by pull request volume.
| If you are... | Start with |
|---|---|
| An open-source maintainer | repowise bot or CodeRabbit (both free on public repos) |
| A small team wanting an AI reader on every PR | CodeRabbit, or repowise AI review for the lowest per-PR cost |
| Large codebase, cross-file bugs matter | Greptile |
| Already on GitHub Copilot Business or Enterprise | Copilot code review |
| Already writing code in Cursor | Cursor Bugbot |
| A missed defect costs far more than any review | Claude Code Review on the riskiest PRs |
| Already using a quality platform | Codacy |
| Want repeatable comments and a merge gate | repowise bot, or Reviewdog on top of your linters |
| Code cannot leave your network | Reviewdog, or Greptile or CodeRabbit Enterprise |
Scroll the table sideways to see every column.
A simple rule works well: use an AI reviewer for judgment and a deterministic bot for repeatability. For fourteen AI reviewers compared as products, including GitLab and Bitbucket support, see best AI code review tools. For org-wide installs on private code, see repowise for teams.
FAQ
- What is a PR bot?
- A PR bot is a GitHub App or CI job that reads each pull request and posts comments or status checks. Some run linters, some use an LLM to write a review, and some, like the repowise bot, post facts from an index of the repository such as which callers a change breaks and which files usually change together.
- What is the best PR review bot for GitHub?
- For hosted AI review, CodeRabbit and Greptile lead. If you already pay for Copilot, try GitHub Copilot code review first, and Cursor teams should try Bugbot. For repeatable, rule-based comments, use the repowise bot or Reviewdog. Codacy suits teams that want quality scans and PR feedback in one platform.
- What AI reviews my GitHub pull requests automatically?
- CodeRabbit, Greptile, GitHub Copilot code review, Cursor Bugbot, Claude Code Review, Codacy's AI Reviewer and repowise AI review all review GitHub pull requests automatically once installed or enabled. Copilot needs an automatic-review rule on the repository; the GitHub Apps review every new PR and push by default or by setting.
- How much do GitHub PR review bots cost?
- As of November 2026, the repowise bot is the lowest-cost hosted reviewer: Signals is free on every repository, and its AI review costs about 15 cents a pull request on average, about $15 per 100 pull requests. CodeRabbit starts at $24 per developer per month, Greptile Pro charges $1 per credit with no seat fee, Codacy Team is $18 per developer per month, a Bugbot run averages $1.00 to $1.50 and Claude Code Review about $15 to $25. Reviewdog is free.
- Are there free PR review bots for open source?
- Yes. CodeRabbit reviews public repositories free, Greptile is free for qualified non-commercial projects with an OSI-approved license, Codacy is free for open-source projects, Reviewdog is open source, and the repowise bot is free on public repos. See free AI code review for open source for the full list.
- How do I add a PR bot to a GitHub repository?
- Most PR bots install as a GitHub App. Open the bot's install page, choose the organization and select the repositories it may read; the next pull request on those repositories gets a review. CI-based bots such as Reviewdog run as a step in a GitHub Actions workflow and post comments with the workflow's token.
- Which PR review bot is best for self-hosting?
- Reviewdog runs entirely in your own CI. Greptile and CodeRabbit offer self-hosting on their Enterprise plans. The repowise bot is hosted, and its open-source engine runs coverage, doc-drift, security and change-risk gates as a GitHub Action in your own CI.
- Should I choose deterministic or LLM-based review?
- Use deterministic review for repeatable findings tied to rules or repository facts, and LLM review for an explanation of intent and cross-file impact. Many teams run both, with the deterministic layer acting as the gate.



