Remote approvals, policy checks, and execution evidence for unattended AI agents.

Let AI agents investigate and operate infrastructure through declared actions, bounded by policy and host-side checks. Approvals when required, with an audit trail. By Protectorate.