Skip to content

How does human-in-the-loop work?

Configure per-agent approval gates; when triggered, agents pause and notify humans via Slack or the dashboard.

Updated Apr 18, 20263 viewsapprovalshitlgovernance

Human-in-the-loop approvals

Configure approval gates per agent. When an agent reaches a gate, it pauses and requests human approval.

Configuration examples

  • Require approval before deploying code to production.
  • Require approval before LLM calls above $1.
  • Require approval before reading customer data.

Approval flow

  1. Agent reaches a gate.
  2. Approver gets notified via Slack or the dashboard.
  3. Approver approves, rejects, or requests changes.
  4. Everything is logged in an immutable audit trail (365-day retention).

Full runbook: see the Approvals page for policy examples.

Was this helpful?

Still need help?

Our team typically responds within one business day.

Submit a ticket →
How does human-in-the-loop work? — Dobby Help