Last updated on Jun 22, 2026

Tags:

• Human-in-the-Loop
• Governance
• Audit

Options

Open in ChatGPTAsk ChatGPT about this page↗Open in ClaudeAsk Claude about this page↗

Approvals

The Approvals page is the organization-wide Human-in-the-Loop (HITL) queue. Review and act on pending approval requests.

Access from the sidebar by clicking Approvals. The badge shows the number of pending requests.

Real-Time Updates

The page updates in real-time. A green indicator shows "Real-time Updates" / "Live" status. Click Refresh to manually reload.

Stats Cards

Five metrics across the top:

Metric	Description
Pending	Approval requests awaiting review (orange)
Approved Today	Requests approved in the last 24h (green)
Rejected Today	Requests rejected in the last 24h (red)
Expired Today	Requests that reached the 24h hard expiry (gray)
Avg Response Time	Average time to respond with trend indicator

SLA Summary

A summary bar shows SLA performance:

• Within - Percentage of approvals resolved within SLA target (green)
• At Risk - Approaching SLA deadline (orange)
• Breached - Exceeded SLA target (red)
• SLA Target - Responsiveness goal for SLA reporting (Within / At Risk / Breached). Default: 5 min. Missing the SLA target does not expire the request.
• Hard expiry - Every approval auto-expires 24 hours after it is raised, independent of the SLA target. On expiry the operation does not proceed (treated as denied) and the rule's on_timeout behavior (block or halt) applies. The UI shows a live countdown to this 24h deadline.

Click Analytics to view detailed SLA analytics.

Approval Queue

Status Filter

Use the status dropdown to filter the queue:

• Pending - Requests awaiting action
• Approved - Recently approved requests
• Rejected - Recently rejected requests
• Expired - Requests that reached the 24h hard expiry

Approval Cards

Each pending approval shows:

Field	Description
Agent	Agent name, icon, and Trust Tier badge
Operation	What the agent is attempting
Session	Link to session for context
Time Pending	How long it's been waiting
SLA Status	Within SLA, At Risk, or Breached
Priority	Low, Medium, High, or Critical

SLA Indicators

Each card shows SLA status:

• Green border - Within SLA target
• Orange border - At risk (>80% of SLA time elapsed)
• Red border - SLA breached

Approval Actions

For each pending request:

Action	Description
View Agent	Open the agent detail page in a new tab
Details	View full details of the approval request
Reject	Block operation (requires reason)
Approve	Allow operation to proceed

Processing Approvals

Approving

1. Review the operation details
2. Click Approve
3. Optionally add a comment
4. The operation resumes

Result:

• Activity retries and succeeds
• Event logged in audit trail
• Trust score unchanged

Rejecting

1. Review the operation details
2. Click Reject
3. Enter a rejection reason (required)
4. Confirm rejection

Result:

• Operation is rejected
• Counts as a rule trigger; a trust penalty applies only if the rule's threshold is exceeded in the rolling window (no flat per-event penalty)
• Event logged in audit trail

Expiry

Two independent clocks apply to every approval:

• SLA target (default 5 min) — a responsiveness goal for reporting only. Missing it moves the request to At Risk / Breached but does not expire it.
• Hard expiry (24h) — if no decision is made within 24 hours, the request auto-expires:
  • The operation does not proceed — expiry is treated as a denial, and the rule's on_timeout setting (block or halt) is applied.
  • The request appears in the "Expired Today" stat and under the "Expired" status filter.

Expiry carries no flat trust penalty. Like any denied outcome it counts as one rule trigger; a trust penalty applies only if that rule's trigger count exceeds its threshold within the rolling window (see Trust Scores → Recovery).

Filtering

Filter the queue by:

• Priority - Low, Medium, High, Critical
• Agent - Specific agent
• Team - Owning team
• SLA Status - Within, At Risk, Breached

Next Steps

1. Organization Settings - Manage teams, roles, and organization configuration
2. Monitor Sessions - Review the full session context with Session Replay before approving or rejecting