clauderabbit

🐰 ClaudeRabbit

Self-contained, agent-native AI code review. A free, private, CLI-less replacement for CodeRabbit that finds bugs, security issues, performance problems, and quality risks in your code — using your coding agent's own model, against your local git diff.

No subscription. No login. No external API. Your code never leaves your machine.

---

Why ClaudeRabbit?

CodeRabbit is great, but it runs a separate CLI that uploads your diffs to a third-party service and requires an account/subscription. ClaudeRabbit does the same job using the model your coding agent is already running:

	CodeRabbit	ClaudeRabbit
Review engine	External CLI → CodeRabbit API	Your agent's own model
Account / subscription	Required	None
Code leaves your machine	Yes (diffs uploaded)	No
Extra install	coderabbit CLI	None — just the skill
Works in	Supported agents	Any agent that loads SKILL.md

ClaudeRabbit only ever runs git (always) and gh (for the optional PR workflows).

Quickstart

Once installed, just tell your agent:

Review my code

or run the slash command (Claude Code):

/clauderabbit:review

Abilities

Ability	Slash command	What it does	Triggers (natural language)
review	/clauderabbit:review	Review local git changes; findings grouped 🔴 Critical / 🟡 Warning / 🔵 Info with file:line and concrete fixes.	"review my code", "check for bugs", "security review", "run clauderabbit"
autofix	/clauderabbit:autofix	Apply fixes one at a time with per-change approval, then a single consolidated commit + verification re-review.	"fix the findings", "apply fixes", "clauderabbit autofix"
pr-review	/clauderabbit:pr-review	Review a GitHub PR and optionally post a summary review + inline comments via gh.	"review this PR", "review pull request 42"
summary	/clauderabbit:summary	CodeRabbit-style walkthrough: changed-files table, overview, impact notes, test plan — for a PR description.	"summarize my changes", "pr walkthrough"

Review scopes

/clauderabbit:review                      # all changes (default)
/clauderabbit:review uncommitted          # working tree + staged
/clauderabbit:review staged               # staged only
/clauderabbit:review --base main          # this branch vs. main
/clauderabbit:review --base-commit abc123 # vs. a specific commit

Installation

Option A — Claude Code plugin (recommended)

Run these three commands inside Claude Code:

/plugin marketplace add https://github.com/TacoGod900/clauderabbit.git
/plugin install clauderabbit
/reload-plugins

That's it. You now have the /clauderabbit:* slash commands, the skills, and the clauderabbit-reviewer subagent. Verify by typing /clauderabbit: — you should see review, autofix, pr-review, and summary.

Use the full https://….git URL (not the TacoGod900/clauderabbit shorthand). The shorthand can default to SSH and fail with Host key verification failed if your machine hasn't connected to GitHub over SSH before. The HTTPS URL always works for a public repo.

Already added the marketplace and want the latest version? Run /plugin marketplace update clauderabbit, then /reload-plugins.

Option B — Portable skills (any coding agent)

ClaudeRabbit's skills/ folder is portable SKILL.md — copy it into your agent's skills directory. Pick the row for your agent:

Agent	Project path	Global path
Claude Code	.claude/skills/	~/.claude/skills/
Cursor	.cursor/skills/	~/.cursor/skills/
Codex	.codex/skills/	~/.codex/skills/
Gemini CLI	.gemini/skills/	~/.gemini/skills/
GitHub Copilot	.github/skills/	~/.copilot/skills/
Cline	.cline/skills/	~/.cline/skills/
Continue	.continue/skills/	~/.continue/skills/
Roo Code	.roo/skills/	~/.roo/skills/
Kilo Code	.kilocode/skills/	~/.kilocode/skills/
Windsurf	.windsurf/skills/	~/.codeium/windsurf/skills/
OpenCode	.opencode/skills/	~/.config/opencode/skills/
Goose	.goose/skills/	~/.config/goose/skills/
Amp / Kimi CLI	.agents/skills/	~/.config/agents/skills/

For example, to install globally for Claude Code:

# macOS / Linux
cp -r skills/* ~/.claude/skills/

# Windows (PowerShell)
Copy-Item -Recurse skills\* "$env:USERPROFILE\.claude\skills\"

Then start a fresh session and say "review my code." Skills trigger by natural language, so you don't need slash-command support in your agent.