How does the human-in-the-loop approval process work?

Every tool has one of three permission levels:

• Allowed — executes automatically
• Requires approval — pauses for human confirmation
• Disabled — never runs

You configure these at the org level with overrides per tool, per server OS, or per connection type.

When approval is required:

• Notification sent to your configured channel (Slack, Teams, PagerDuty, or SMS) with the tool name, VixPro AI's reasoning, server context, and what it's trying to accomplish
• Authorization happens in the dashboard with MFA — never via Slack reply or SMS response. This prevents approval-by-SMS attacks where a spoofed message could authorize a destructive operation

For scheduled maintenance, approvals can be pre-authorized at schedule creation time so overnight runs don't require a 3am confirmation.