ai-code-review-agent
CrashBytes/ai-code-review-agent
Production-ready AI code review automation using Claude API and GitHub Actions
Instalação
npx convez-ai CrashBytes/ai-code-review-agent1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
README
A tradução para o português deste README ainda está na fila.
AI Code Review Agent
Production-ready AI code review automation using Claude API and GitHub Actions. Automatically analyzes pull requests for security vulnerabilities, code quality issues, and architectural concerns.
Features
- Automated code review on every pull request
- Security vulnerability detection (SQL injection, XSS, auth bypasses)
- Code quality analysis (complexity, duplication, naming)
- Architectural feedback (SOLID principles, coupling, scalability)
- Performance issue identification (N+1 queries, inefficient algorithms)
- Error handling validation (try-catch, promise rejections)
- Configurable severity thresholds
- Smart finding deduplication
- Cost-effective token management
- Production monitoring and metrics
Quick Start
1. Install in Your Repository
cd your-repo
npm install @anthropic-ai/sdk @octokit/rest
2. Add GitHub Actions Workflow
Create .github/workflows/ai-code-review.yml:
name: AI Code Review
on:
pull_request:
types: [opened, synchronize, reopened]
permissions:
contents: read
pull-requests: write
jobs:
review:
runs-on: ubuntu-latest
timeout-minutes: 10
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- uses: actions/setup-node@v4
with:
node-version: '18'
cache: 'npm'
- run: npm ci
- name: Run AI code review
env:
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
run: npx ai-code-review
3. Configure Secrets
- Get your Anthropic API key from https://console.anthropic.com
- Go to your repository Settings → Secrets and variables → Actions
- Add
ANTHROPIC_API_KEYwith your API key GITHUB_TOKENis automatically provided
4. Customize Configuration (Optional)
Create .ai-review-config.json:
{
"model": "claude-sonnet-4-20250514",
"minSeverity": "medium",
"reviewFocus": [
"security-vulnerabilities",
"code-quality",
"architectural-concerns",
"performance-issues",
"error-handling"
],
"excludePatterns": [
"**/node_modules/**",
"**/dist/**",
"**/build/**"
]
}
Configuration Options
| Option | Type | Default | Description |
|---|---|---|---|
model | string | claude-sonnet-4-20250514 | Claude model to use |
maxTokens | number | 4000 | Maximum tokens per response |
temperature | number | 0.0 | Model temperature (0 = deterministic) |
minSeverity | string | medium | Minimum severity: low, medium, high, critical |
reviewFocus | array | All areas | Review focus areas |
excludePatterns | array | Common artifacts | File patterns to exclude |
Review Focus Areas
security-vulnerabilities: SQL injection, XSS, authentication bypassescode-quality: Duplication, complexity, naming, magic numbersarchitectural-concerns: SOLID violations, coupling, abstractionsperformance-issues: N+1 queries, inefficient algorithmserror-handling: Missing try-catch, unhandled promises
Cost Management
Typical costs for a 100-developer team:
- 50 pull requests/day
- Average 10 files per PR
- Approximately $75/month in API costs
- $30/month in GitHub Actions usage
- Total: ~$105/month
ROI: 190:1 return based on time savings
Example Output
The agent posts comments directly on pull requests:
### AI Code Review 🟠 HIGH
**Category**: security
**Issue**: SQL injection vulnerability due to string concatenation in database query
**Recommendation**: Use parameterized queries with prepared statements to prevent SQL injection attacks
**Example Fix**:
```javascript
// Instead of:
const query = `SELECT * FROM users WHERE id = ${userId}`;
// Use:
const query = 'SELECT * FROM users WHERE id = ?';
db.query(query, [userId]);
Finding ID: a1b2c3d4e5f6g7h8 Generated by AI Code Review Agent
## Advanced Usage
### Local Testing
```bash
export ANTHROPIC_API_KEY="your-key"
export GITHUB_TOKEN="your-token"
export GITHUB_EVENT_PATH="path/to/test-event.json"
node src/review-agent.js
Custom Severity Thresholds
{
"minSeverity": "high"
}
Only critical and high severity findings will be posted.
Language-Specific Rules
{
"fileTypeRules": {
"typescript": {
"requireTypes": true,
"noExplicitAny": true
},
"python": {
"checkTypeHints": true
}
}
}
Development
Running Tests
npm test
Linting
npm run lint
Architecture
GitHub PR → Webhook → GitHub Actions → Review Agent
↓
Claude API Analysis
↓
Finding Deduplication
↓
GitHub PR Comments
Security
- API keys stored in GitHub Secrets
- No code execution, only static analysis
- Path traversal protection
- File size limits
- Input sanitization
Troubleshooting
No comments appear
- Check GitHub Actions logs for errors
- Verify
ANTHROPIC_API_KEYis configured - Ensure workflow has
pull-requests: writepermission - Check file exclusion patterns aren't too broad
422 errors in logs
GitHub API returns 422 when comment line numbers don't exist in the diff. This is normal for files with extensive changes. The agent logs warnings but continues.
High costs
- Reduce
maxTokensin configuration - Add more exclusion patterns
- Increase
minSeveritythreshold - Review file chunking strategy
Contributing
Contributions welcome! See CONTRIBUTING.md
License
MIT License - see LICENSE
Support
- Tutorial: https://crashbytes.com/articles/tutorial-production-ai-code-review-agent-claude-github-actions-automated-analysis-2025
- Issues: https://github.com/CrashBytes/ai-code-review-agent/issues
- Discussions: https://github.com/CrashBytes/ai-code-review-agent/discussions
Related Projects
Built with Claude API by CrashBytes