Skip to main content
Windows only. Screenshots of the actual notifications are being finalized.
When the Desktop Agent acts on something you send to an AI tool, it can show a notification so you understand what happened. Each notification carries a per-item message that comes from the policy that matched.

​
Masked

Some sensitive content was replaced with placeholder tokens (for example a phone number became [PHONE_NUMBER_1]) before being sent. The AI sees only the masked version, and masking cannot be overridden. The masking always happens, but the notification itself is optional — your administrator can suppress the mask toast for specific endpoints (typically high-frequency ones, such as inline autocomplete, where a toast on every keystroke would be noisy). If you don’t see a notification, masking may still have been applied.

​
Blocked

The request was stopped before reaching the AI service — nothing left your machine. The notification explains the reason (for example, a disallowed topic or sensitive data). You cannot override a block.
A block can also come from an administrator action — for example an active Kill Switch, or an Agent that hasn’t been assigned to a project yet. If everything is unexpectedly blocked, contact your administrator.

​
What administrators see

Every mask and block is recorded as a trace in Opticon monitoring, so your administrator can see which policy applied — useful when troubleshooting.