Agentic cross-file context
The reviewer reads callers, definitions, and related files on demand — no vector DB, no indexing — so it catches broken callers and contract violations a diff-only pass misses.
Prowl Suite · Code Review
prowl-review posts a walkthrough summary, inline findings with committable fixes, and an @prowl-review chat bot on every pull request — using multi-pass specialists, cross-file context, and linter/SAST grounding. BYOK, zero hosting, no rate limits.
Not another diff-only reviewer
The reviewer reads callers, definitions, and related files on demand — no vector DB, no indexing — so it catches broken callers and contract violations a diff-only pass misses.
Parallel correctness, security, performance, and test lenses are merged and de-duplicated by a judge pass into one clean, ranked result — optionally across multiple providers at once.
ESLint, Ruff, Gitleaks, Semgrep, and osv-scanner feed real signals into the review, so deterministic issues are caught and hallucinations are curbed.
A skeptical second pass plus confidence scoring and a severity threshold keep the output high-signal — fewer nitpicks, more real bugs.
Bring your own Claude, OpenAI, or Gemini API key. It never leaves your runner.
Inference goes straight to your provider — no proxy, no prowl-review server, no telemetry.
You pay the provider directly, so we never meter usage. Per-review cost stays in cents.
Get started in two steps
Add your provider key as the PROWL_AI_KEY secret, then drop in the workflow. Open a PR and prowl-review posts a review.
# .github/workflows/prowl-review.yml
name: prowl-review
on:
pull_request:
types: [opened, synchronize, ready_for_review, reopened]
permissions:
pull-requests: write
issues: write
checks: write
contents: read
jobs:
review:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: prowl-tools/prowl-code-review@v1
with:
ai-key: ${{ secrets.PROWL_AI_KEY }}Free forever, BYOK. Made for agents, controlled by humans.