Scan your codebase.
Generate AI guardrails.
viberails reads your existing project — framework, structure, naming, tests — and generates rules that AI tools, hooks, and CI actually follow.
how it works
Reads the room before it sets rules
Framework, tooling, structure, naming, tests. It pulls defaults from the codebase you already trust instead of inventing a policy on top of it.
Checks your setup before you configure
Validates test runner, hook manager, linter, and typecheck before you spend time configuring. Missing tools can be installed inline or skipped.
Generates rules AI tools actually follow
Scan results become .viberails/context.md — natural-language directives wired into Claude Code hooks, pre-commit checks, and CI. Not a template. Derived from your code.
Enforces on every surface
Pre-commit hooks, Claude Code integration, and GitHub Actions. Violations are caught during AI edits, on commit, and in pull requests.
where it shows up
Claude Code PostToolUse hook validates files as they're written. Catches violations mid-session, not after.
Pre-commit hook runs check --staged. Optional typecheck and lint hooks included.
GitHub Actions workflow runs check --enforce to block merges with violations.
get started
$ npx viberails $ pnpm dlx viberails enforced rules from scan
AI-readable directives
raw detection data
Next.js, React, Vue, Angular, Svelte, Astro, Remix, and 15+ more frameworks. ESLint, Biome, Prettier. Vitest, Jest, Playwright, Cypress. npm, pnpm, yarn, bun.