Agent Test Evidence
Operate the deterministic evaluation harness: edit and extend a versioned golden dataset, run per-agent release decisions against the full agent registry with agents refused rather than passed when they carry no evidence, keep a run history that makes a regression an attributable outcome change, and export a machine-readable evidence bundle the release gate reproduces.
Sign in to continue
Intelligence and Transform assets require a verified email. Use Google or a one-time link sent to your Gmail, other personal, or work email. Your cached reports remain available.
or
We record signed-in asset use and financial-data deliveries to measure adoption and data-provider licensing. How we use this information