CodeRabbit pricing, free plan and limits

Last updated: 2026-10-11

CodeRabbit reports 6M repositories and 17K customers for its AI code reviewer, which comments on pull requests across GitHub, GitLab, Azure DevOps and Bitbucket. It also reviews unmerged work inside VS Code, Cursor and Windsurf and from a command-line tool, using language models from OpenAI and Anthropic, with one-click fixes for the problems it finds.

About CodeRabbit

CodeRabbit is an AI code reviewer from CodeRabbit Inc., a San Francisco company founded in 2023 by Harjot Gill, its chief executive, and Gur Singh. Scale Venture Partners led a $60M Series B in September 2025, with NVentures, the venture arm of NVIDIA, taking part, and a $143M Series C at a $1.5B valuation, co-led by Atomico and Smash Capital, followed on August 12, 2026. The product reads each pull request, writes a plain-language summary, leaves line-by-line comments and offers one-click fixes, then keeps learning from how reviewers respond.

It works where the code is already hosted: GitHub, GitHub Enterprise Server, GitLab, self-managed GitLab, Azure DevOps and both Bitbucket editions. Before a pull request exists, the same reviewer runs as an extension in VS Code and the Windsurf editor, and inside the Cursor editor, or from a command-line tool. Under the hood it calls language models from OpenAI and Anthropic and runs third-party linters and security scanners in sandboxes. Around that core sit Triage, which orders a queue of open pull requests, and Change Stack, which splits a large diff into readable layers.

The audience is any team whose pull request queue has outgrown its senior reviewers, which now often means teams merging code written by agents, such as the autonomous engineer Devin, Claude Code or the agent mode in GitHub Copilot. Open-source maintainers use it free on public repositories, which is why it sits in HokAI's free-tier tools list. Teams steer it with a .coderabbit.yaml file or the web settings, including path-based instructions that can hold one folder to stricter rules than another, plus a pre-merge checks list and a spending cap for metered reviews. It is a first pass, not an approver: a person still decides what merges, and the accuracy figures in the pros and cons explain why.

Rivals split by pricing shape and host support. Greptile bills by review credit. Qodo sells a flat team fee. Cursor sells a review add-on called Bugbot. HokAI files all of them under AI coding assistants, part of the code development group. The guide Best AI Code Review Tools in 2026 compares the field on cost, data handling and where each one loses, and Best AI Coding Assistants in 2026 covers the editor side. Teams with private code can also check which tools carry SOC 2 attestations before shortlisting.

Pricing

Four plans, quoted per developer per month. Essentials: $30 monthly, $24 on annual billing. Team: $60 monthly, $48 annual.

Advanced: $72 on annual billing. Enterprise is quote-only and adds SSO, audit logs and self-hosting. Public repositories are free on the open-source plan, and new accounts get a trial before any charge.

Only developers who open pull requests count as seats. 40 per agent minute. The vendor renamed Pro to Essentials and Pro Plus to Team without changing prices, so older third-party listings may use the old names.

For scale, Greptile Pro is $30 per seat and GitHub Copilot Business is $19 per seat.

Key Features

• Pull request reviews: Posts a plain-language summary, a walkthrough with diagrams and line-by-line comments on each eligible pull request, with one-click fixes for unresolved findings on every plan.
• IDE and CLI reviews: Reviews uncommitted changes in VS Code, Cursor and Windsurf, or from a coderabbit review command with install paths for macOS, Linux and Windows.
• Linters and scanners: Runs 50+ third-party linters and security tools, such as ESLint and Ruff, in sandboxed environments and invokes only the ones a repository's files need.
• Learnings and chat: Stores reviewer corrections as editable learnings, applies them to later reviews, and answers questions in chat on the pull request.
• Pre-merge checks and slop detection: Enforces quality gates before merge, including a check that flags low-quality AI-generated pull requests, on every plan.
• Triage and Change Stack: Team and above rank a cross-repository queue of open pull requests by value and risk and post it to Slack, while Change Stack breaks a large diff into navigable layers.
• MCP and web context: Pulls context from up to 5 MCP servers on Essentials (20 on Enterprise) and from live web search, which is on by default.
• Security review: Advanced adds continuous monitoring that reports severity, exploitability and reachability; Deep Scan full-repository scans are a separate on-demand product.

Pros

• Martian's open Code Review Bench, read on 2026-10-11, scores CodeRabbit at 61.8% F1, with 71.1% precision and 54.6% recall, and lists 335,229 pull requests for it. The benchmark is open source and scored from public pull requests, which makes it a better signal than a vendor's own test.
• A team split across GitHub, Azure DevOps or a self-managed GitLab can run a single reviewer on all of it instead of buying a second tool for each host.
• Public repositories get reviews free with no expiry, so an open-source maintainer can judge the tool on real pull requests before any private code is connected.
• The vendor's security page says it is SOC 2 Type II audited annually and that nothing is stored after a review, and its data processing addendum is published, which shortens a procurement checklist.
• Buyers rate it 4.4 out of 5 on G2 across 144 reviews and 4.2 on Gartner Peer Insights from 21, as read on 2026-10-11.

Cons

• Free only on public repositories. Private code needs a paid plan once the 14-day trial ends, so a solo developer has no permanent free option, unlike Greptile's one-developer Starter plan.
• Included reviews are capped per developer per hour (5 on Essentials, 12 on Enterprise). Past the cap, usage billing meters extra reviews at $0.25 per reviewed file, so bursts of agent-written pull requests can push the bill above the seat price unless an admin sets a spending cap.
• The entry plan links one repository and leaves out test generation, Triage and post-merge actions, which start on Team; blast-radius analysis and security monitoring need Advanced.
• The benchmark's recall figure is just over half, so many fixes developers later made were never flagged. Treat the comments as a first pass, not an approval.
• It reviews every eligible pull request on the default branch automatically and skips drafts, so teams that want a quieter feed must write path instructions or pause reviews after a set number of commits.
• Self-hosting and company-selected data regions are Enterprise-only, so teams that must keep code inside their own network have to go through sales.

Visit CodeRabbit Official Website