Modes: read-only and changes

The AI agent has three modes — pick one by the task and by how far you trust it.

Read-only

The agent reads files and logs, explains code and proposes a fix, but changes nothing itself. The safe mode for understanding a problem.

WITH CONFIRMATION (the default)

The agent prepares an edit and shows it to you; it is applied only after your yes. Right for everyday work: you see every change before it happens.

Autonomous

The agent applies what can be undone — file edits — on its own. Convenient for long, repetitive tasks.

Even here there is a line it does not cross: panel actions with no way back — deleting a domain, a database, a mailbox — always ask you, autonomous mode included. What cannot be reversed stays with a human.

What holds in every mode

• A backup is taken before a file is edited; any change can be rolled back in one click.
• Every action the agent takes is written to a log that cannot be altered or erased — it is open to you on the agent's activity page.
• The agent is locked to your own account and cannot see anyone else's.
• The agent has no access to the wallet: it cannot pay, order or renew anything.

How to switch

AI settings in the service's file manager.

Recommendations

Start with read-only to see what the agent suggests. Keep confirmation mode for real work. Take a full service backup before large changes.