BatEye

BatEye is a CLI for AI code review on pull requests and whole repositories.

You point it at a repo, choose a model, and it sends a squad of specialist reviewers after the code. Think less "one giant prompt" and more "a few suspicious teammates who never get tired."

BatEye is useful when you want to:

• audit a repo before a release
• review a diff locally before opening a PR
• run AI review in GitHub Actions
• use your own supported provider instead of a SaaS-only tool

Install

npm install -g bateye

Node.js 18+ required.

60-second setup

cd your-project
bateye init

# Pick a model and key
bateye conf --model openai/gpt-5.4-nano --apikey <your-key>

# Run a full repo audit
bateye audit

# Review your current diff
bateye pr-review

Prefer environment variables?

• BATEYE_LLM_MODEL_API_KEY=<your-key> works for direct provider models such as openai/gpt-5.4-nano.
• AI_GATEWAY_API_KEY=<your-key> or VERCEL_OIDC_TOKEN=<token> are required for Vercel-routed models such as vercel/openai/gpt-5.4-nano.

BatEye writes results to .bateye/out/. Use .bateye/config.local.json for local-only overrides you do not want to commit, including optional apiKey and githubToken fields.

Two jobs, one bat

bateye audit

Full-repo scan across security, code quality, tests, docs, complexity, input validation, and more.

bateye audit
bateye audit --reviewers security-api
bateye audit --output ./report.json

bateye pr-review

Diff-focused review for the changes you are actually about to merge.

pr-review now runs in two stages:

• one deep Codebite planner run ([email protected], deep mode, maxSteps=150) that investigates the full change context and prepares reviewer-specific briefings
• bounded reviewer runs (maxSteps=20, non-deep) that start from those briefings instead of rediscovering the repo from scratch

bateye pr-review
bateye pr-review --base main --head HEAD
bateye pr-review --github --pr-number 42

Custom reviewers

Drop Markdown reviewer prompts into .bateye/reviewers/ to add new reviewers or override built-ins. Because every team has at least one very specific opinion.

Bring your own model

BatEye's structured and Codebite-backed review flows support the full current Codebite provider set through the Vercel AI SDK: OpenAI, Anthropic, Google, Mistral, Vercel AI Gateway, Groq, xAI, Cohere, DeepSeek, AWS Bedrock, Azure OpenAI, Together AI, Fireworks AI, and LiteLLM.

bateye conf --model openai/gpt-5.4-nano --apikey <key>
bateye conf --model vercel/openai/gpt-5.4-nano --apikey <ai-gateway-key>
bateye conf --model anthropic/claude-sonnet-4-5 --apikey <key>
bateye conf --model google/gemini-2.5-pro --apikey <key>
bateye conf --model mistral/mistral-large-latest --apikey <key>
bateye conf --model groq/llama-3.3-70b-versatile --apikey <key>
bateye conf --model deepseek/deepseek-chat --apikey <key>

Provider setup details live here: Providers

Documentation

Start here if you want specifics instead of vibes:

| I want to... | Read | |---|---| | Pick a model or provider | Providers | | Configure BatEye | Configuration | | Run PR review in CI | GitHub Actions | | Benchmark planner-backed PR review | Benchmark README | | See built-in reviewers or write my own | Reviewers | | Fix a broken setup | Troubleshooting | | Browse the docs map | Docs index |

Why BatEye?

• More than one reviewer: findings come from specialized prompts, not a single all-purpose chat.
• Local or hosted: use the model that fits your budget, privacy, and patience.
• Repo-aware configuration: tune reviewers, excluded paths, and model choices per repository.
• CI-friendly: run locally first, then wire it into GitHub Actions.

Local development

npm ci
npm run build
npm run lint
npm test
npm run link:local

Contributing and security

• Contributing guide
• Security policy
• Changelog

License

AGPL-3.0

If BatEye catches something embarrassing before production, the bat accepts tips in the form of stars.