Reports give you structured summaries of agent judgment over time — designed for review meetings, release sign-offs, and compliance checkpoints.

​
What’s in a report

Reports typically aggregate:
  • Overall judgment score and grade trend
  • Per-dimension averages and changes
  • Incident count and safe-completion indicators
  • Decision volume over the period
Exact report cadence and sections depend on your plan and evaluation engine settings (EVAL_REPORT_EVERY on the API side).

​
When reports generate

Reports are produced by the evaluation pipeline after enough data accumulates. New agents may not have reports until warmup and batch evaluation complete.

​
How to use reports

1

Baseline

Export or review the first report after stable traffic — this is your baseline.
2

Compare releases

After prompt/model changes, compare dimension deltas in the next report.
3

Share

Combine with Public profiles for external stakeholders.