Juror AI is a lightweight, multi‑model PR bot that runs entirely inside your own GitHub Actions runner, so your code never leaves the repository. Add a single workflow file to .github/workflows/juror.yml, drop in your provider keys (e.g., OpenAI or Anthropic), and open a pull request. Juror instantly posts a sticky "Juror is reviewing…" comment, then replaces it with findings, a merge confidence score, and a bill.

Each model receives the diff, runs in its own sandbox, and reports findings that are deduplicated through a five‑stage lossless merge. The bot defaults to high recall, but you can switch to consensus mode for higher precision. You pay per model call, not per seat, and the total cost—often as low as $0.91 for 3 models—appears in the receipt.

Juror doesn’t require an index or SaaS; it uses the repo’s own read/search tools, keeping the code local. It’s not an autofix bot, linter, or chat interface—just a diff‑reviewer that posts inline comments in a single batched review. The workflow file is optional; you can also run the binary locally with ``npx juror-ai review --pr 1234``.

Overall, Juror delivers multi‑model coverage, cost transparency, and no external indexing, making it a superior alternative to single‑model bots.