One API call adds EU AI Act Article 14 compliant human oversight to any AI agent — across every framework, rail, and action type.
Agent hits a protected action and calls POST /v1/pauses. Returns immediately with a pause_id.
LoopPause routes a structured request to Slack or email with full action context. The human approves or rejects — no code required.
LoopPause fires an HMAC-SHA256 signed webhook to your agent. The proof is independently verifiable offline and satisfies Article 14 audit requirements.
Request
POST /v1/pauses
Authorization: Bearer sk_live_...
{
"agent_id": "billing-agent",
"action": {
"type": "payment",
"description": "Pay Acme Corp $12,450"
},
"recipients": [{
"channel": "slack",
"target": "#finance-approvals"
}],
"webhook_url": "https://acme.com/hooks"
}Response
HTTP 202 Accepted
{
"pause_id": "clx9f2a1b0000...",
"status": "pending"
}
// Webhook fires with signed proof:
{
"decision": "approved",
"human_response": true,
"signature": "sha256=a3f9..."
}Works with LangGraph, CrewAI, AutoGen, n8n, or plain HTTP. No SDK required — one POST request, any language.
Agents never proceed without a valid signed proof. Returns 503 on unavailability — never a silent pass-through. The proof is verifiable offline.
HMAC-SHA256 signed authorization records on every decision. Non-repudiable, auditor-ready. A technical demonstration of what EU AI Act compliance infrastructure looks like.
LoopPause is currently in early access for developers building production AI agents. If you're working on agent governance, EU AI Act compliance, or just want human oversight that doesn't block your process — reach out to join the waitlist.
Request early accesshello@looppause.com — no forms, no funnels