Juror runs parallel pull request reviews using multiple frontier models through native agent harnesses to increase bug detection recall. It uses code-aware similarity and a referee mechanism to deduplicate findings from different providers. The tool executes on existing GitHub Actions runners to keep code within the local environment.