CodeRabbit

What it does

Writes code

YesA cloud Coding Agent clones the repo into a sandbox and implements a typed request, but it is in open beta and each starting point is enabled separately per organisation.Coding Agent is the part of CodeRabbit that writes code. It takes a request — typed by you, or raised by a CodeRabbit review finding, a failing…docs.coderabbit.ai/code ↗

Debugs and fixes

YesIt reads failing CI logs on the pull request and posts inline fixes on the offending lines, and a Fix CI finishing touch can land those fixes as a commit or stacked PR.When a CI/CD pipeline fails on a pull request, CodeRabbit reads the failure output and posts inline comments with suggested fixes — directly on the…docs.coderabbit.ai/pr-reviews/cicd-pipe… ↗

Writes tests

PartlyUnit test generation is a Team-and-above finishing touch, absent from the entry Essentials plan, and is itself labelled open beta on GitHub, GitLab and Azure DevOps.Everything in Essentials, plus tasks and actions upstream and downstream of the review process: Triage, CodeRabbit Plan, issue planning, unit test…docs.coderabbit.ai/management/plans ↗

Reviews code

YesReviewing diffs written by other people or agents is the core product: every pull request is reviewed automatically on open, with manual commands for older PRs.Once installed, CodeRabbit automatically triggers a review when a pull request is opened against the main branch of any repository.docs.coderabbit.ai/faq ↗

Whole-codebase work

PartlyReviews reach beyond one repo through linked-repository analysis, but the number of linked repos is capped per plan (1 on Essentials) and automatic linking needs Team or above.The number of linked repositories you can configure depends on your plan. Automatic repository linking is available on Team, Advanced, and Enterprise.docs.coderabbit.ai/knowledge-base/multi… ↗

Ships it

PartlyAccepted work lands in Git as a commit on the PR branch or a stacked pull request, pushed with your own credentials; nothing in the docs deploys or releases the change.Delivery is always a separate, explicit step. A task that finishes successfully has changed nothing in your repository until you accept its patch and…docs.coderabbit.ai/code/deliver-changes ↗

How much it does unattended

Runs autonomously

PartlyReviews run unattended on every PR, but the coding agent stops at a proposed patch by default and a person must accept it before anything reaches the repository.It proposes, you decide. By default a finished task produces a proposed patch: a diff you read, comment on, revise, reject, or accept. Delivery is a…docs.coderabbit.ai/code ↗

Multi-agent

YesThe vendor describes several specialised agents running in parallel inside one review, rather than a single prompt, plus separate chat and finishing-touch agents.Specialized AI agents working in parallel: Review, Verification, Chat, Pre-Merge Checks, and Finishing Touchesdocs.coderabbit.ai/overview/architecture ↗

Agent permissions

PartlyThe agent inherits your Git permissions and branch protections and cannot push around them, and Plan mode adds an approval gate, but there is no per-action policy engine documented.Delivery uses your Git provider authorization, so your branch protections and permissions apply exactly as they would to a push you made yourself.docs.coderabbit.ai/code ↗

Which models it runs on

Claude

YesAnthropic is named as a model provider that receives code for reviewing, though CodeRabbit selects the models and buyers cannot pick a specific Claude version on hosted plans.Your code is shared with OpenAI and/or Anthropic for reviewing purposes onlydocs.coderabbit.ai/faq ↗

GPT

YesOpenAI is named alongside Anthropic as a provider that processes code during reviews; the choice between them is CodeRabbit's, not the customer's, on hosted plans.Your code is shared with OpenAI and/or Anthropic for reviewing purposes onlydocs.coderabbit.ai/faq ↗

Open models

NoThe hosted service names only OpenAI and Anthropic as inference providers; no open-weight option is published for CodeRabbit Cloud.Your code is shared with OpenAI and/or Anthropic for reviewing purposes onlydocs.coderabbit.ai/faq ↗

Model choice

PartlyHosted plans give no model picker, but Enterprise self-hosted deployments may point CodeRabbit at their own LLM provider or account for inference.Bring your own model — connect CodeRabbit to your own large language model provider or account.docs.coderabbit.ai/self-hosted/overview ↗

Where you use it

In your editor

YesA VS Code extension reviews uncommitted changes locally and works in Cursor, Windsurf and other VS Code-compatible editors, with chat, docstrings and test generation excluded.Review code changes before you commit and create agent-ready Coding Plans, directly in your IDE with the CodeRabbit extension.docs.coderabbit.ai/ide ↗

On the command line

YesA terminal tool reviews local Git changes before commit and emits structured JSON for coding agents, though the CLI carries an Open Beta badge.The CodeRabbit CLI analyzes local Git changes using the same pattern recognition that powers our PR reviews.docs.coderabbit.ai/cli ↗

In your pipeline

YesIt acts on pull requests automatically and reads pipeline output from five CI systems; an agentic API key also authenticates the CLI inside CI runners and bots.This works across multiple CI/CD platforms: GitHub Actions, GitLab CI/CD, CircleCI, Jenkins, and Azure DevOps Pipelines.docs.coderabbit.ai/pr-reviews/cicd-pipe… ↗

In a browser

YesA hosted dashboard at app.coderabbit.ai carries the coding agent, Triage queue, settings, reports and billing, separate from the Git platform surfaces.Find it in the sidebar under Code, or go directly to app.coderabbit.ai/code.docs.coderabbit.ai/code ↗

Whose machine it runs on

Self-hosted

PartlyA container image runs the review agent inside your own network, but only for Enterprise customers at 500 seats or more, arranged through sales during onboarding.The self-hosted option is available for CodeRabbit Enterprise customers with 500 or more user seats.docs.coderabbit.ai/self-hosted/overview ↗

Not established: Open source

What it costs to run

How it meters

YesBilling is a per-developer seat plus optional metering: over-limit reviews at a fixed price per reviewed file, and cloud agent work billed by the minute.Continued reviews cost $0.25 per reviewed file and are added to the monthly invoice.coderabbit.ai/pricing ↗

Free tier

YesA standing Free plan needs no card and covers public and private repos, and every paid plan opens with a 14-day trial that also requires no card.Unlimited public and private repositories, no credit card required.docs.coderabbit.ai/management/plans ↗

API access

PartlyThe REST API for metrics, users, roles, audit logs and security findings is Enterprise-only; lower tiers get an agentic API key for headless CLI review authentication.The CodeRabbit API provides programmatic access to review data, AI Deep Scan findings, and administrative operations.docs.coderabbit.ai/api ↗

MCP server

NoCodeRabbit consumes MCP servers for review context but publishes none of its own, so another agent cannot query it over MCP; plan tiers cap the connections at 5 to 20.CodeRabbit as MCP client: CodeRabbit acts as the MCP client, not the server. It ingests data from your connected MCP servers.docs.coderabbit.ai/connections/mcp-serv… ↗

Bring your own key

PartlyBring-your-own inference exists only in the Enterprise self-hosted deployment, where the agent sends prompts to your own LLM account; hosted plans use CodeRabbit's providers.The agent sends review prompts, source code, and related data to your configured large language model provider for inference.docs.coderabbit.ai/self-hosted/overview ↗

Buying it for a team

A company can buy it

YesThe subscription is bought by an organisation against its Git provider org, with seats, admin and billing-admin roles, and AWS, GCP and Claude marketplace billing on Enterprise.CodeRabbit ingests users from your git provider (GitHub, GitLab, Bitbucket, and Azure DevOps) and assigns seats based on your organization's active…docs.coderabbit.ai/management/seat-assi… ↗

Seat model

YesSeats are per developer and only developers who open pull requests consume one; seats are reassignable, and auto-approval reclaims a seat unused for 30 days.No, you will only be charged for developers who create pull requests. You can also manually assign purchased seats to specific developers.coderabbit.ai/pricing ↗

Pooled budget

PartlyReview allowances are per developer per hour, not pooled, while pay-as-you-go overage spending runs through one organisation-wide path with a single monthly cap.Your organization shares one pay-as-you-go billing path across all PR and CLI reviews.docs.coderabbit.ai/management/usage-bas… ↗

Admin controls

YesAdmins choose what happens past the review limit, set spending caps and seat modes, and scope which repositories and connections CodeRabbit may touch; custom RBAC is Enterprise.After the included review limit is reached, an admin chooses what happens next with the After included review limits setting in the Usage-based…docs.coderabbit.ai/management/usage-bas… ↗

Audit log

PartlyA searchable, exportable administrative audit log covering seats, roles, API keys and config exists, but only for Enterprise organisations and only for admin-level roles.The Audit Logs give Enterprise organizations a searchable record of administrative actions taken across workspace settings, billing controls, seat…docs.coderabbit.ai/management/audit-logs ↗

Single sign-on

PartlySAML 2.0 single sign-on with Okta or Entra ID, plus SCIM provisioning, is an Enterprise-plan capability set up with the account team rather than self-serve.CodeRabbit supports Okta, Microsoft Entra ID, and any SAML 2.0-compliant identity provider.docs.coderabbit.ai/management/sso ↗

What happens to your code

Opt out of training

YesNo opt-out is needed because customer code is never used for training by CodeRabbit or its providers; separately, retention of learnings and model-output logging can be switched off.CodeRabbit never uses customer code for training. This remains true whether data retention is enabled or disabled.docs.coderabbit.ai/faq ↗

Data residency

PartlyAn EU SaaS deployment is listed only under Enterprise; the trust centre lists US subprocessors and the CLI defaults to a US region, so ordinary plans are US-hosted.coderabbit.ai/pricing ↗

Getting out

PartlyReview metrics export as CSV capped at 10,000 rows, with a larger Metrics API on Enterprise only; account deletion is irreversible and immediately cancels the subscription.This page covers the Dashboard CSV export. For programmatic access, use the Metrics Data API (Enterprise plan required).docs.coderabbit.ai/guides/data-export ↗

Certifications

YesThe trust centre publishes SOC 2 Type 2, ISO 27001:2022 and GDPR, with SOC 3, pentest and zero-data-retention amendments from OpenAI and Anthropic available on request.Compliance SOC 2 TYPE 2 ISO 27001:2022 GDPRtrust.coderabbit.ai ↗