Audit

Each decision can be traced and reviewed.

Agent step record.: Each step an agent takes is recorded with its prompt, the context it used and the person who asked. Records are signed, kept for the period you set and exportable to your security monitoring.
Run traces.: Each run is traced across the services it touches, with its sources and cost.
Evaluation.: Agents are tested before they are published, each draft can be checked against red-team prompts, and recorded runs can be evaluated again after a change.
Review queues.: Knowledge the platform extracts is queued for your team to confirm, and conflicting facts are queued for a person to resolve.
Change control.: Agent versions are kept, and a change can be rolled back.
Retention and erasure.: You set the retention periods. Memories can be deleted on request, and expired items are removed automatically.
Continuity.: Backups are taken and restores are tested.