\n\n# crev — **C**ode **REV**iew\n\n### Use your existing subscriptions\n\nAI code review tools typically require separate API keys and per-token billing. You're already paying for Claude Code, Gemini CLI, Codex, or other AI coding subscriptions -- crev runs reviews through those CLI tools directly. No extra API keys. No extra billing.\n\n### Run multiple reviewers in parallel\n\nDefine a review schema with multiple reviewers using different models and prompts, and crev executes them all simultaneously. A security expert on Claude, an architect on Gemini, and a performance reviewer on Codex -- all running at the same time, all finishing in the time it takes the slowest one to complete.\n\n### Configurable and reusable\n\nYAML schemas define your review pipeline. Inline prompts or external agent files shape what each reviewer focuses on. Run the same review across PRs, branches, or commits. Share schemas across your team.\n\n```bash\n# 1 reviewer, fast feedback\ncrev run --schema quick --base main\n\n# 3 reviewers + triage pass\ncrev run --schema standard --base main\n\n# 6 specialized reviewers + dedup triage\ncrev run --schema thorough --base main\n\n# Full codebase analysis\ncrev run --schema security --analyze\n```\n\n## Install\n\n```bash\nnpm install -g @caiokf/crev\n```\n\nOr with Homebrew:\n\n```bash\nbrew install caiokf/tap/crev\n```\n\nOr use without installing:\n\n```bash\nnpx @caiokf/crev init\nnpx @caiokf/crev run --schema quick --base main\n```\n\n## Quick Start\n\n```bash\n# Initialize crev in your project\ncrev init\n\n# Run a quick review against main branch\ncrev run --schema quick --base main\n\n# Review a specific PR\ncrev run --schema quick --pr 42\n\n# Review uncommitted changes\ncrev run --schema quick --type uncommitted\n\n# Full codebase analysis (no diff)\ncrev run --schema security --analyze\n\n# Check runtime health\ncrev doctor\n\n# Smoke test all runtimes\ncrev doctor --ping\n```\n\n## How It Works\n\n1. **Define a schema** -- Pick which runtimes, models, and prompts to use\n2. **crev generates the diff** -- From your branch, PR, or commit\n3. **Reviewers run in parallel** -- Each runtime gets the diff and reads files from the codebase\n4. **Output is normalized** -- All results are merged into a single JSON review\n5. **Optional triage** -- A devil's advocate AI pass deduplicates and prioritizes findings\n\n```yaml\n# .crev/schemas/standard.yaml\ndescription: Three-reviewer standard check\nreviewers:\n - name: Engineer\n runtime: claude\n model: sonnet\n prompt: >\n You are a senior software engineer reviewing code changes.\n Focus on correctness, error handling, edge cases, and readability.\n\n - name: Security\n runtime: gemini\n model: gemini-2.5-pro\n prompt: >\n You are a senior security engineer reviewing code changes.\n Focus on injection vulnerabilities, auth flaws, and secrets exposure.\n\n - name: Architect\n runtime: codex\n model: gpt-5.4\n prompt: >\n You are a senior software architect reviewing code changes.\n Focus on coupling, abstraction quality, and scalability concerns.\n\ntriage:\n enabled: true\n runtime: claude\n model: opus\n```\n\nEach reviewer can use either an inline `prompt` or an `agent` field pointing to any external file:\n\n```yaml\nreviewers:\n - name: Security\n runtime: claude\n model: opus\n agent: .crev/agents/security-reviewer.md\n```\n\n## Supported Runtimes\n\n| Runtime | CLI | Auth |\n|---|---|---|\n| Claude Code | `claude` | Subscription or `ANTHROPIC_API_KEY` |\n| Codex | `codex` | `OPENAI_API_KEY` |\n| Copilot | `copilot` | `gh auth login` |\n| Gemini CLI | `gemini` | `GEMINI_API_KEY` or `GOOGLE_API_KEY` |\n| Kimi | `kimi` | `MOONSHOT_API_KEY` |\n| CodeRabbit | `cr` | `cr auth status` |\n| OpenCode | `opencode` | `~/.local/share/opencode/auth.json` |\n| Droid | `droid` | Subscription |\n| MastraCode | `mastracode` | Subscription |\n| Pi | `pi` | Subscription |\n\nRun `crev doctor` to check which runtimes are available. Add `--ping` to verify end-to-end functionality.\n\n## Commands\n\n| Command | Description |\n|---|---|\n| `crev init` | Interactive setup |\n| `crev run --schema ` | Execute a review |\n| `crev run --analyze` | Full codebase analysis (no diff) |\n| `crev doctor [--ping]` | Check runtime health (optionally run ping tests) |\n| `crev stats --schema ` | Reviewer effectiveness stats |\n| `crev config [--layers]` | Show resolved configuration |\n| `crev list` | List schemas and runtimes |\n| `crev show [file]` | Pretty-print a review file (default: latest) |\n| `crev fix [file]` | Auto-fix review issues using AI agents |\n| `crev triage [file]` | Run triage on an existing review |\n| `crev diff` | Preview diff that would be reviewed |\n| `crev schema show ` | Display schema details |\n| `crev schema validate` | Validate schemas |\n| `crev schema init ` | Scaffold a new schema |\n| `crev update` | Regenerate AI tool skills |\n| `crev help ` | Detailed help (run, schema) |\n\n## Project Structure\n\nAfter `crev init`:\n\n```\n.crev/\n├── config.yaml # Global config and model aliases\n├── schemas/\n│ ├── quick.yaml # 1 reviewer, fast\n│ ├── standard.yaml # 3 reviewers + triage\n│ └── thorough.yaml # 6 reviewers + triage\n└── reviews/ # Review output (JSON)\n```\n\n## Related\n\n- **[valet](https://github.com/caiokf/valet)** -- The standalone runtime adapter library used by crev\n\n## License\n\nMIT\n","readmeFilename":"README.md"}
![\"crev](\".github/banner.svg\")