Today, tool approval for Port AI is configured per user, and tools that can run actions, such as run_action and trigger_run, run without approval by default.

Admins should be able to set an organization-wide approval policy for Port AI tools, for example requiring approval for every tool that runs actions, that individual users cannot override.

This would let security-conscious organizations make sure no action is executed by Port AI without a human approving it, without relying on each user's personal settings.