How do Doozy approval gates work?

Safe defaults (what you get out of the box)

Every new Doozy agent starts with approval required on external actions. That means: sending messages, making purchases, deleting things, changing production, modifying permissions, and exporting sensitive data all pause and wait for your OK. Internal work like research, drafting, and organizing runs freely.

You see the exact text, the recipient, the evidence behind it, and whether you can undo it. You approve, reject, or send it back with feedback. Approving covers that one action, not every similar one in the future.

Custom per agent

You can tighten or loosen the rules per agent. A sales agent might be allowed to update CRM fields on its own but still need approval for customer emails. A research agent might run fully autonomous because it never touches anything external. A finance agent might require approval on everything.

The rules layer: workspace defaults set the floor, and per-agent settings can only be stricter than the workspace. An agent can't talk its way past an admin rule no matter what the request says.

YOLO mode (full autonomy)

For agents you fully trust on specific categories, you can create always-allow rules that let the agent act without pausing. This is the "YOLO" path: the agent drafts and sends, updates and files, without stopping for you.

The key: YOLO is opt-in per category, not a global switch. You can YOLO one agent on CRM updates while keeping approval on emails. And you can revoke it anytime — the agent goes back to asking.