Approve or reject an agent action
For operators who want to keep control over every sensitive action triggered by an agent (write, command, external tool call).
Manage tool permissions
For operators who want to review what each agent may do on this machine, and withdraw a consent, without waiting for an action to raise a new request.
Inspect a tool
For operators who want to know what a tool actually is before authorising it: what it takes as input, what it gives back, what it requires, and which credentials are stored against it.