FeatureOctober 10, 20261 min read
Run evals as specific users and tenants
Assign user and tenant identities to eval cases or suite defaults so you can test agents whose behavior depends on who is asking.
evalsidentitymulti-tenancy
Test an identity-aware agent under different users and tenants in the same eval suite. Set an identity on an individual case, or define defaults for the suite and override them where needed.
The eval runner applies that identity through the same admission and tool-context path as a normal execution. Tools that use signed identity assertions receive the eval case's subject.
Use this to check personalized responses and tenant-scoped behavior, including agents that require an identity before they run. These are trusted, organization-authored test identities, not stored visitor login tokens.