Skip to main content
Sensitive agent actions (cards, domains, KYC, formation, connecting services) can require human approval. When gated, the originating endpoint returns 202 with a pending-approval body instead of executing: 
{
  "status": "pending_approval",
  "approval_id": "uuid",
  "action": "cards.create",
  "primitive": "cards",
  "title": "Issue virtual card \"Ads\"",
  "message": "This action requires human approval before it executes."
}
Approve it (the API replays the frozen action) or deny it via the endpoints below. Two mounts:
  • Per-user: /v1/users/:user_id/approvals (the user’s queue)
  • Cross-user: /v1/approvals (operator dashboard, all users)
MethodPathPurpose
GET/approvals?status=List approvals
GET/approvals/:idGet one
POST/approvals/:id/approveApprove & execute
POST/approvals/:id/denyDeny
Which actions are gated is configured per primitive on the Account Kit (requiresApproval). See Architecture → Approvals for the execution model.