#KiteBBW $KITE
Card payments protect the buyer and expose the sellerđ.
A customer can dispute a charge months later, and the merchant eats the chargeback.
Kite's service provider guide describes a flow that flips this, and I think the flip is the most underrated design choice in the agent economyđ.
The flow is short.
An agent calls an endpoint, and the service replies with a 402 Payment Required that states the price, the pay-to address and a timeout.
The agent answers with a signed authorization drawn from the user's approved session.
The service verifies it, a facilitator settles the transfer on chain, and only then does the service deliver the response.
Settlement FIRST, delivery SECONDđ€·ââïž.
The merchant doesn't extend credit or wait for an invoice to clear.
That's what makes tiny per-request prices workable in principleđ€.
Card rails carry fixed costs per transaction that small payments can't absorb, and per-request pricing only works if the payment itself is nearly free.
The guide says any service that already speaks x402 or the machine payments protocol Stripe co-authored can be paid by a Passport agent today.
More than 90 providers were integrated at mainnet launch on April 30, 2026.
But the risk didn't disappear, it moved to the buyer.
Once settlement clears before delivery, a service that returns junk has already been paid, and the provider guide I read describes no refund or dispute step.
The user's protection is the sessionđ: the per-transaction cap and the total cap limit what any one bad call can cost
In my view those caps do the job chargebacks do in the card world, just before the damage instead of afterđ
That trade looks fine at small per-call prices, real trouble is big buys of the agent.
If settlement is final before the service responds and a paid call returns something useless, what should replace the chargeback: reputation scores for services, refund escrow built into the protocol, or nothing, because per-call prices are too small to justify the machinery?
Drop 'kite' if you got good info here
Card payments protect the buyer and expose the sellerđ.
A customer can dispute a charge months later, and the merchant eats the chargeback.
Kite's service provider guide describes a flow that flips this, and I think the flip is the most underrated design choice in the agent economyđ.
The flow is short.
An agent calls an endpoint, and the service replies with a 402 Payment Required that states the price, the pay-to address and a timeout.
The agent answers with a signed authorization drawn from the user's approved session.
The service verifies it, a facilitator settles the transfer on chain, and only then does the service deliver the response.
Settlement FIRST, delivery SECONDđ€·ââïž.
The merchant doesn't extend credit or wait for an invoice to clear.
That's what makes tiny per-request prices workable in principleđ€.
Card rails carry fixed costs per transaction that small payments can't absorb, and per-request pricing only works if the payment itself is nearly free.
The guide says any service that already speaks x402 or the machine payments protocol Stripe co-authored can be paid by a Passport agent today.
More than 90 providers were integrated at mainnet launch on April 30, 2026.
But the risk didn't disappear, it moved to the buyer.
Once settlement clears before delivery, a service that returns junk has already been paid, and the provider guide I read describes no refund or dispute step.
The user's protection is the sessionđ: the per-transaction cap and the total cap limit what any one bad call can cost
In my view those caps do the job chargebacks do in the card world, just before the damage instead of afterđ
That trade looks fine at small per-call prices, real trouble is big buys of the agent.
If settlement is final before the service responds and a paid call returns something useless, what should replace the chargeback: reputation scores for services, refund escrow built into the protocol, or nothing, because per-call prices are too small to justify the machinery?
Drop 'kite' if you got good info here