PluginInstalls cleanly

Get a second review of your code

CodeRabbit Skills and Plugins

CodeRabbit's official plugin lets your assistant ask CodeRabbit to review your changes for bugs and security problems, then helps you fix what it finds.

Works with
Claude Code, Codex CLI, Gemini CLI
Good for
Security testing, Web developers, Build an app

What it does

CodeRabbit is a code review service. It reads changes to your code and points out bugs, security problems and risky spots before you keep them. This is CodeRabbit's own plugin. It lets your assistant send your latest changes to CodeRabbit and bring the findings back into your chat.

It works with the CodeRabbit CLI, a small program you install once and sign in to. Install it with brew install coderabbit, or with CodeRabbit's install script. CodeRabbit's own skill prefers a package manager such as Homebrew, because the script runs code from the internet. Then sign in with coderabbit auth login (EU accounts: add --region eu). The skill does not start the sign-in for you. If you are not signed in, it asks you to run that command.

The plugin adds two skills (instruction files):

  • code-review runs the review. Once installed, it is the default review skill, so a plain "review my code" sends your changes to CodeRabbit.
  • autofix collects the open review comments on a pull request and fixes them one by one with your approval. It uses the GitHub CLI (gh). After you approve, it commits the fixes and can push, create a pull request or post a comment.

The versions differ by assistant:

  • Claude Code gets version 1.1.1 from Anthropic's directory. It adds one review agent and a /coderabbit:coderabbit-review command. That command pre-approves coderabbit, cr and git commands.
  • Gemini CLI gets the newer, unreleased version from the main branch (--ref main). It has two agents, one of them a read-only autofix helper, and a /coderabbit:review command.

The newer features in the repo are still marked "Unreleased" in its changelog.

Each review sends your code changes and the project context needed to review them to CodeRabbit's service (CLI network docs).

We installed the Claude Code plugin in our sandbox. We did not run the CLI install lines, the Codex line or the Gemini CLI line there, and we did not run a review.

Who it is for

People who let an AI assistant write code and want a second reviewer to check it. The assistant that wrote the code is not the best judge of its own work.

Example prompts

  • "Review my code."
  • "Run a CodeRabbit review on the changes in this folder only."
  • "Fix the issues CodeRabbit found, one at a time, and ask me before each change."
  • "How many CodeRabbit reviews do I have left right now?"

How to install

brew install coderabbit
curl -fsSL https://cli.coderabbit.ai/install.sh | sh
/plugin marketplace update
/plugin install coderabbit@claude-plugins-official
npx skills add coderabbitai/skills -a codex
gemini extensions install https://github.com/coderabbitai/skills --ref main

How to use it

Install the CodeRabbit CLI first (brew or the install script), sign in with coderabbit auth login (EU accounts: add --region eu), then tell your assistant: Review my code.

Check it works

  1. Run cr doctor in your terminal. The CLI docs say it checks the install, sign-in, your Git folder and the connection to CodeRabbit.
  2. In Claude Code, run /plugin and look for coderabbit.
  3. For Gemini CLI, run gemini extensions list (README).

Common problems

  • coderabbit: command not found right after install — no exact error text — start a new terminal window, or run the reload command the installer prints (CLI docs).
  • Sign-in goes to the wrong region for EU users — no exact error text — skip the installer's browser sign-in and sign in later with coderabbit auth login --region eu (CLI docs).
  • A review stops with an offer to use credits — "awaiting_confirmation" — you have used the reviews included for this hour. Say yes only if you want to pay for this one (cli-workflows.md, CLI docs).

What to watch out for

  • Each review sends your code changes and the project context needed to review them to CodeRabbit's service (README, Data and Privacy; CLI network docs)
  • Once installed it becomes the default review skill, so a plain request such as review my code goes to CodeRabbit (SKILL.md)
  • Reviews have rolling hourly limits: the Free plan allows 3 CLI reviews per developer per hour (CodeRabbit plans docs); above the limit a paid organisation can pay per review at $0.25 a reviewed file (usage-based add-on docs), and the CLI shows the file count and maximum price and spends nothing until you agree (CLI docs)
  • The Claude Code install is version 1.1.1 from Anthropic's directory; its review command pre-approves coderabbit, cr and git commands (commands/coderabbit-review.md). The newer features in the repo are still marked Unreleased
  • The autofix skill uses the GitHub CLI (gh): after you approve, it commits the fixes and can push, create a pull request or post a comment (autofix SKILL.md)
  • The CLI install script runs a script from the internet; CodeRabbit's own skill prefers a package manager such as Homebrew
  • The Gemini CLI line installs unreleased code from the main branch, a newer version than the Claude Code one (README)

Licence and author

Licence
MIT
Author
CodeRabbit
Source code
github.com/coderabbitai/skills

How we tested it

We installed it on an empty test computer on 17 Oct 2026. Installed and started in a clean test system. Not yet used in a real AI session.

Test level

Not run: "curl -fsSL https://cli.coderabbit.ai/install.sh | sh" (not a form the sandbox runs). Not run: "/plugin marketplace update" (not a form the sandbox runs). Not run: "gemini extensions install https://github.com/coderabbitai/skills --ref main" (not a form the sandbox runs). Not yet used on real work.

Technical details of the test
Date
17 Oct 2026
Method
Intake sandbox run on 2026-10-17 in a clean container with no account: cloned the repository and checked its skill, agent and plugin files, ran "npx -y skills --help"; installed coderabbit@claude-plugins-official with the Claude Code command line.

Checked again on 17 Oct 2026

Alternatives

  • PR Review Toolkit agents — review agents that run inside Claude Code, with no outside service.
  • Trail of Bits skills — security-focused audit skills.
  • OpenQodex — runs open-source scanners on your own computer.

FAQ

Does it cost money? The review limit is per hour, on a rolling window. CodeRabbit's Free plan allows 3 CLI reviews per developer per hour, with up to 150 files per review. Paid plans start at $24 per developer per month, billed yearly (CodeRabbit plans docs). Paid organisations can also turn on pay-per-review at $0.25 a reviewed file (usage-based add-on docs). The CLI shows the price first, and only developers with an assigned seat can approve it (CLI docs).

Does my code leave my computer? Yes. The CLI sends the changes under review to CodeRabbit's service (README, CLI network docs).

Does it work with Gemini CLI? Yes, as a pre-release extension installed from the main branch. That is newer, unreleased code (README).

Can it change my code by itself? The autofix skill asks before each change and before it pushes. The Claude Code version has no separate read-only helper.

Sources