Getting started
Prerequisites#
- A GitHub organization where you can install applications
- At least one repository with pull request activity
- Product specs, architecture decisions, or context docs in markdown (you can add these after installing)
No CI changes, no YAML pipeline edits, no build step.
Step 1 — Install the GitHub App#
Add Prelint to your GitHub organization at app.prelint.com and select which repositories to review. New organizations start with $10 in credits and no credit card.
Step 2 — Add context to your repository#
Drop product specs, decision logs, or architecture docs into the repository. Markdown files are indexed automatically, so put them wherever makes sense for your project — specs/, docs/, or the repository root.
A spec does not need to be long. This is enough to review against:
# Authentication
- Users authenticate via email/password or Google OAuth.
- Sessions expire after 30 days of inactivity.
- Failed login attempts are rate-limited to 5 per minute.
- Password reset tokens expire after 1 hour.Step 3 — Open a pull request#
Prelint reviews every pull request within seconds. Findings appear as inline comments pinned to the exact lines that triggered them, plus a check run summary with an overview.
Expected output on a PR that contradicts a spec:
Conflicts with existing spec —
specs/pricing/settlements.mdrequires charges stored in vendor currency. This PR stores in customer currency, creating FX risk on advance bookings. Suggestion: store in vendor currency, convert for display only.
Step 4 — Act on the findings#
Each finding names what does not match, which spec it contradicts, and a suggested fix. Push a correction and Prelint re-reviews — findings from earlier reviews on the same pull request resolve automatically.
Next steps#
- Read How reviews work to understand what happens between the webhook and the comment
- Tune what gets reviewed in Configuration reference
- If AI agents open your pull requests, set up the loop in Working with AI coding agents
- Back to the documentation home