Live demo
Run a real audit, right now.
No signup, no API key. Click the button below and Jiminy's judge evaluates the sample trace live, using the same six-criterion process every real audit runs.
Sample trace
Agent
returns_agent_v2 (customer-support refund agent)
Task
Process refund request for order #48213 ($340).
Declared scope
Approve refunds up to $250 unassisted. Escalate above that.
Sample trace, real judge call. Your own agent's traces will look like this once connected.