How to add automated code review to GitHub in 10 minutes
Connect an AI reviewer to your GitHub repos without changing your workflow. Every pull request gets checked automatically. No CI/CD setup required.
Most developers spend 20 to 40 minutes reviewing a single pull request. For solo developers or small teams, this time often gets skipped entirely — which means bugs and security issues ship directly to production.
There is a faster way: let an AI do the first pass. This guide shows you how to connect CodeSentinel to your GitHub repositories so every pull request gets an automatic review before any human looks at it.
What this looks like in practice
You push code. You open a pull request. Within a minute, CodeSentinel posts inline comments on the specific lines with issues: security vulnerabilities, performance problems, logic errors. You see the feedback right in GitHub, exactly like comments from a human reviewer. You fix what needs fixing. You merge.
What you need
- A GitHub account with admin access to your repositories
- A server or VPS to run the service
- An Anthropic API key for the AI analysis
- A PostgreSQL database (NeonDB free tier works)
Step 1: Create a GitHub App
Go to GitHub Settings → Developer Settings → GitHub Apps → New GitHub App. Set the webhook URL to your server. Grant read access to pull requests and code. Generate a private key. This takes about 3 minutes.
Step 2: Deploy CodeSentinel
Clone the repository. Create a .env file with your credentials:
DATABASE_URL=postgresql://your-neon-db-url
GITHUB_APP_ID=your_app_id
GITHUB_APP_PRIVATE_KEY=your_private_key
GITHUB_WEBHOOK_SECRET=your_webhook_secret
ANTHROPIC_API_KEY=sk-ant-your-key
NEXTAUTH_URL=https://your-domain.com
NEXTAUTH_SECRET=any-random-stringRun npx prisma db push, then npm run build && npm start.
Step 3: Install on your repositories
Go to your GitHub App settings and install it on the repositories you want covered. From this point, every pull request triggers an automatic analysis.
What gets reviewed
By default, CodeSentinel checks for security vulnerabilities, performance issues, code quality problems, and architectural inconsistencies. Each finding includes the exact file, line number, severity, and a suggested fix.
