Prelint
On teams running several AI reviewers, about 40% of the issues fixed before merge are caught by Prelint.
Three things Prelint does#
Reviews intent, not implementation. Every pull request is checked against your specs, compliance rules, business constraints, and tooling decisions. If code violates what the company decided, Prelint flags it inline on the exact lines that triggered it.
Closes the loop for AI agents. Findings arrive as standard GitHub review comments. Claude Code, Cursor, and Copilot read them, fix the issues, and push again — no human in the middle for each iteration.
Records the decision. Who reviewed, what was found, how it was resolved, and when. When an auditor asks for oversight of AI-written code, you export the record instead of reconstructing it.
Get started in five minutes#
Install the GitHub App, drop your product specs into the repository, and open a pull request. Prelint reviews it within seconds. Start at Getting started.
Where to go next#
- Getting started — install the app and get your first review
- How reviews work — the five-stage review pipeline
- Product drift — the six ways AI code drifts from your product
- Working with AI coding agents — the agent feedback loop
- Configuration reference — every
prelint.jsonproperty - Pricing and credits — $1 per completed review
- Security — what Prelint stores, sends, logs, and deletes
- Research: AI Code Pulse — 56,706 pull requests analyzed
- FAQ — short answers to the common questions