🦇 BatEye

Free, open-source AI code auditor and PR reviewer. BatEye sends a squad of specialized AI agents through your codebase - each one echolocating a different class of problem - then delivers a structured report with severity scores, file references, and actionable fixes.

Why a squad instead of just asking your AI coding assistant? A single AI conversation misses things. BatEye runs parallel, purpose-built reviewers - security, complexity, test coverage, docs, validation, and more - each one deeply focused on its domain. You get coverage that a general chat session simply can't match.

Run an audit. Walk into your boss's office. "Found 47 issues. Fixed them all. Saved us $30k in external audit fees." Promotion secured. 🦇

Install

npm install -g bateye

Free and open source (AGPL-3.0). Use your own models - Ollama, LM Studio, or any local/cloud provider - to keep costs lower. See providers →

Requires Node.js 18+

Quick start

# 1. Get an API key (Anthropic, OpenAI, OpenRouter, etc.) - or use a local model
export BATEYE_LLM_MODEL_API_KEY=your-key

# 2. Initialize in your repo
cd your-project
bateye init

# 3. Audit your codebase
bateye audit

# 4. Review your latest changes
bateye pr-review

Audit results → .bateye/out/audit.json · PR review results → .bateye/out/pr-review.json 🦇

Two modes, one tool

bateye audit - full codebase scan

Deploys all reviewers across your entire codebase. Use it before releases, after refactors, or when you inherit a new repo and need to know where the bodies are buried.

bateye audit                             # Every reviewer, whole codebase
bateye audit --reviewers security-api    # One reviewer only
bateye audit --output ./report.json      # Custom output path

bateye pr-review - diff-focused review

Focuses reviewers on the diff only - what changed, what broke, what was missed. Run it locally or wire it into GitHub Actions so every PR gets reviewed automatically.

bateye pr-review                         # Local diff (origin/main...HEAD)
bateye pr-review --base main --head HEAD
bateye pr-review --github --pr-number 42 # Post inline comments to a GitHub PR

→ GitHub Actions setup - trigger on /review comment, on every push, or on a schedule.

Reduce PR review waiting time. Enable auto-approve and BatEye will approve the PR automatically when no significant issues are found - your team stops waiting on a human for straightforward changes:

{
  "prReview": {
    "autoApprove": { "enabled": true, "maxSeverity": "low" }
  }
}

Configuration

All behaviour is controlled by .bateye/config.json in your repo root. bateye init creates it, bateye config set edits it:

{
  "model": "anthropic/claude-sonnet-4-5",
  "exclude": ["generated", "migrations"],
  "disabledReviewers": {
    "audit": ["accessibility"],
    "prReview": ["inline-docs"]
  }
}

Pick a model, exclude noisy paths, disable reviewers that don't apply to your stack. Full config reference →

How it works

BatEye runs parallel, purpose-built AI reviewers - not a single chat prompt, but a coordinated squad:

| Reviewer | Looks for | |---|---| | security-api | Injection flaws, secrets exposure, auth gaps | | code-quality | Smells, complexity, maintainability | | test-coverage | Missing tests, edge cases, brittle assertions | | documentation | Missing docs, stale comments | | complexity | Overly complex logic, refactor candidates | | input-validation | Unvalidated inputs, type coercions | | ...and more | Full catalog → |

Each reviewer investigates independently. BatEye deduplicates and synthesizes findings into one report.

You can also write your own reviewers - drop a .md file with a prompt into .bateye/reviewers/ and BatEye picks it up automatically. Override any built-in reviewer or add completely new ones. Custom reviewers →

Supported providers

Use any AI provider - just set BATEYE_LLM_MODEL_API_KEY and pick a model:

bateye config set model anthropic/claude-sonnet-4-5
bateye config set model openai/gpt-4o
bateye config set model ollama/llama3.2        # local, free
bateye config set model openrouter/meta-llama/llama-3.3-70b-instruct

Anthropic · OpenAI · OpenRouter · Google · DeepSeek · Groq · Cerebras · Azure · Ollama (local, free) · LM Studio (local, free) · Vercel AI Gateway · 20+ providers →

Other commands

bateye init                              # Set up .bateye/ in your repo
bateye doctor                            # Verify config, API key, reviewers
bateye models                            # List available AI models
bateye reviewers                         # List all reviewers (built-in + custom)
bateye config show                       # Show current config

Going further

| Topic | Doc | |---|---| | Config fields, schema, model selection | Configuration | | Built-in reviewers, writing custom reviewers | Reviewers | | GitHub Actions - PR review triggers & setup | GitHub Actions | | All supported providers + model formats | Providers | | API keys, environment variables | Configuration → Environment variables | | Something broken | Troubleshooting |

Local development

npm ci
npm run build
npm run link:local       # builds + installs as global `bateye`

npm run test:unit        # fast unit tests
npm run test             # unit + integration

Copy .env.example to .env and fill in your keys for local runs.