AI agents will move money autonomously. Aurel sits between your agent and the payment rail — blocking injected instructions, semantic anomalies, and policy violations before a single transaction executes.
Submit a transaction and watch Aurel evaluate it across three defense layers in real time. Select a scenario or write your own.
The Verification Console requires a minimum viewport of 1024px. Open on a desktop or laptop to use it.
Request API access instead →Three complementary layers between your agent and the payment rail.
Amount thresholds, allowlists & denylists, and velocity limits per agent or workspace — evaluated in microseconds with zero false negatives.
Claude AI reads the agent's reasoning and checks it against the actual transaction. Catches prompt injections and reasoning anomalies that rigid rules miss.
One API call to secure any agentic payment flow. Works with LangChain, CrewAI, AutoGPT, and fully custom agents in minutes.
Send the transaction details and your agent's reasoning. Aurel evaluates deterministic rules first, then runs a semantic analysis with Claude AI, and returns a clear verdict in milliseconds.
allowTransaction is safe to executeflagRequires human review before proceedingblockTransaction must not be executedPOST /api/verify
x-api-key: ig_your_api_key
{
"intent_id": "pay_abc123",
"agent_id": "agent_gpt4",
"amount": 4800.00,
"currency": "USD",
"recipient": "vendor@acme.com",
"agent_context": "User approved a $4,800 invoice
payment to Acme Corp for Q2 services."
}{
"decision": "allow",
"reason": "All rules passed",
"risk_score": 8,
"evaluated_at": "2026-06-22T14:30:00Z",
"intent_id": "pay_abc123"
}Start free during beta. Upgrade as you scale.
Request API access, ask a technical question, or discuss an enterprise deployment. We respond within one business day.
contact@aurel.ioPGP key available on request · Response SLA: 1 business day