Merchants are blocking agents because a scraper and a buyer look identical at the firewall. AGATE answers the question in one free call: signed intent, wallet standing, delivered-work history. Then you decide whether to take the order, take the payment first, or refuse it.
A block stops the scraper and loses the sale, because both arrive the same way. What a merchant actually needs is one bit it can act on: is this agent spending its own settled money, and has anyone paid it for work it delivered? That is what this returns.
POST https://x402-agent-pay.com/agate/verify
Content-Type: application/json
{
"agent": { "wallet": "0x...", "name": "shopper-agent" },
"intent": { "action": "book_appointment",
"target": "bright-angel-lodge",
"amount_usdc": 0.10,
"expires_at": "2026-11-16T20:15:00Z" },
"signature": "0x... EIP-191 personal_sign over the canonical intent string"
}
The signature is checked for free and costs no gas. Canonical string: AGATE1|<action>|<target>|<amount %.6f>|<expires_at epoch>|<wallet lowercase>
{
"decision": "allow_with_limit",
"agent": { "tier": "UNTRUSTED" },
"limits": { "max_usdc": 1.0, "requires_prepayment": true },
"checks": [ { "id": "intent_signature", "status": "pass" }, ... ],
"why": "Intent is sound but the wallet has no delivered-work history ...",
"receipt": { "verify_id": "agate_...", "verify_url": "..." }
}
| Decision | Meaning | What the merchant does |
|---|---|---|
| ALLOW | Signed intent from a wallet with delivered-work history | Serve the request normally, settle after delivery |
| ALLOW WITH LIMIT | Sound intent, thin history, or above the tier ceiling | Take payment first, release the order after it clears |
| DENY | A hard check failed: no signature, bad wallet, sybil flag, expired intent | Refuse, and tell the agent which check failed |
Ceilings: UNTRUSTED 1.00 USDC, STANDARD 5.00, TRUSTED 25.00 per transaction. A refusal is a diagnosis, not a wall: every deny names the check and what would fix it.
# before you fulfil
curl -s -X POST https://x402-agent-pay.com/agate/verify \
-H 'Content-Type: application/json' \
-d '{"agent":{"wallet":"0x..."},"intent":{"action":"book_appointment","target":"my-shop","amount_usdc":0.10}}'
# decision == allow -> fulfil now
# decision == allow_with_limit -> require the x402 payment to clear first
# decision == deny -> 403 with denied_by[]
Two different questions, both free to ask. AGATE answers should I trust this agent before it spends? The facilitator at x402-agent-pay.com/facilitator answers did this payment clear, and can I settle it? AGATE is the check before the money moves, the facilitator is the settlement after. You can use either alone, and they are cheaper together because a refusal never reaches the settlement step.
Spec: /agate/spec · Receipts: GET /agate/verify?receipt=<verify_id> · AgentWorld native copy: agentworld.me/api/agate/verify