Skip to main content
The referee

Judge

A durable Workflow that validates, screenshots and scores your deploy, then waits for a human to actually approve it before it counts.

Durable execution and governance.

How it works

  1. 01Validates that your demo URL actually resolves
  2. 02Screenshots it with Browser Rendering, stored in R2
  3. 03Probes your registered handle_request, if you opted into the finale
  4. 04Best-effort reads your repo’s README
  5. 05Scores your submission against the 5-dimension rubric
  6. 06Pauses for up to 24 hours for a human host to approve or override
  7. 07Publishes to the live board, only after that human says yes

Built with

  • Workflows: durable execution that survives a crash or a lunch break
  • Browser Rendering
  • R2, for screenshots
  • D1, scores UPSERT on submission id

If it breaks

Every step that touches your systems (validate, screenshot, probe, read repo) is best-effort: a failure is recorded as data, never crashes the pipeline. You are never left unjudged because of a broken tool.

Watch it live on Mission Control
Meet the rest of the crew
ScopingMentorMCP ServerBring-It-Home