CodeRabbit

AI-powered code review for pull requests and local changes.

Quick Navigation

Prerequisites Check (MUST RUN BEFORE REVIEW)

Before running CodeRabbit CLI, verify ALL of the following:

# 1. CLI installed?
which coderabbit || echo "MISSING: install with: curl -fsSL https://cli.coderabbit.ai/install.sh | sh"

# 2. Authenticated?
coderabbit auth status 2>&1 | grep -q "Logged in" || echo "MISSING: run coderabbit auth login"

# 3. Git repo has at least one commit? (CRITICAL — CLI crashes with GitError on empty repos)
git rev-parse HEAD >/dev/null 2>&1 || echo "MISSING: repo has no commits — make at least one commit first"

# 4. Base branch exists? (CLI defaults to 'main')
git rev-parse main >/dev/null 2>&1 || echo "WARNING: 'main' branch not found — use --base <branch>"

If any check fails, fix it before running the review. Do NOT proceed with a broken state.

Authentication failure rule: If authentication check fails (step 2), the agent MUST:

1. Stop immediately — do not attempt to run the review
2. Notify the user that CodeRabbit CLI is not authenticated
3. Show the user the exact command to authenticate: coderabbit auth login
4. Wait for the user to complete authentication before retrying
5. Do NOT attempt to run coderabbit auth login on behalf of the user — it requires interactive browser redirect

Quick Start

Run Review

# AI agent workflow (most common) — note: 'review' subcommand is optional
coderabbit review --prompt-only --type uncommitted --no-color

# If base branch is not 'main' (e.g., master, develop):
coderabbit review --prompt-only --type uncommitted --base master --no-color

# Plain text output (human-readable)
coderabbit review --plain --type uncommitted --no-color

Local Capture Script

Persist output to a file for later analysis:

# IMPORTANT: use absolute path to the skill's script directory
python3 ~/.claude/skills/coderabbit/scripts/run_coderabbit.py --output coderabbit-report.txt

Options:

• --output to choose a different file name (saved to .code-review/ in repo root)
• --timeout to adjust the timeout in seconds (default: 1800)
• --base to specify base branch (default: auto-detect from git)

PR Commands

@coderabbitai review          # Incremental review
@coderabbitai full review     # Complete review
@coderabbitai pause           # Stop auto-reviews
@coderabbitai resume          # Resume auto-reviews
@coderabbitai resolve         # Mark comments resolved

Severity Matrix

AI Agent Workflow Pattern

Implement [feature] and then run CodeRabbit CLI in a background terminal.
Wait for it to complete, then read the report. Fix CRITICAL/HIGH issues. Ignore nits.

Step-by-step:

1. Run prerequisites check (see above) — fix any issues before proceeding
2. Detect base branch: git symbolic-ref refs/remotes/origin/HEAD 2>/dev/null or fall back to main/master
3. Run CLI in background: coderabbit review --prompt-only --type uncommitted --base <branch> --no-color
4. Reviews take 7-30+ minutes — run in background (run_in_background=true)
5. Read output when process completes
6. Fix CRITICAL/HIGH findings, skip LOW
7. Limit to 2-3 review iterations maximum

Troubleshooting

[error] stopping cli with no details

Run with DEBUG=* to see the actual error:

DEBUG=* coderabbit review --prompt-only --type uncommitted 2>&1 | grep -E "(ERROR|error|GitError)"

Check the log file:

ls -t ~/.coderabbit/logs/ | head -1 | xargs -I{} cat ~/.coderabbit/logs/{}

Common errors

Check auth status

coderabbit auth status

Linked Repositories (2026-02-18)

CodeRabbit can analyze linked repositories during PR review to catch cross-repo breakages (API/type/dependency drift).

• Configure linked repositories in Knowledge Base settings.
• As of 2026-03-11, Pro plans can link up to 2 repositories for Multi-Repo Analysis.
• Use this when changes in one repo affect contracts in another.
• Treat cross-repo findings as HIGH/CRITICAL when they indicate runtime incompatibility.

Dashboard and Reporting (2026-03-12)

• Dashboard metrics are now split between Git platform reviews and IDE/CLI reviews.
• Reporting surfaces now include Git-platform pages like Knowledge Base, Pre-merge Checks, and Reporting, plus IDE/CLI pages like Summary, Organization Trends, and Data Metrics.
• Team filters are available across dashboards; use them when review volume or findings need to be separated by team rather than repository alone.

Custom Finishing Touch Recipes (Early Access) (2026-02-23)

Define reusable, named "finishing touch" recipes that apply agentic code changes to your PR.

See configuration.md for a minimal example.

Minimal Configuration

# .coderabbit.yaml
language: en-US
reviews:
  profile: chill
  high_level_summary: true
  tools:
    gitleaks:
      enabled: true
    ruff:
      enabled: true

Critical Prohibitions

• Do not introduce fallbacks, mocks, or stubs in production code
• Do not broaden scope beyond what CodeRabbit flagged
• Do not "fix" style nits handled by formatters/linters
• Do not ignore CRITICAL findings; escalate if unclear
• Stop and resolve CLI errors (auth/network) before fixing code
• Do not run CLI on a repo with no commits — it will silently crash

Links

• Documentation
• Changelog
• GitHub
• Schema

Templates

• coderabbit.minimal.yaml — Minimal configuration
• coderabbit.full.yaml — Full example with all options
• agent-prompts.md — Ready-to-use AI agent prompts