AI agents pay us over MPP and Tempo, in production

We have been building toward agents as customers for a while. A few weeks ago an agent paid us a dollar over x402 to provision itself. Since then we wired up MPP, the standard Stripe and Tempo are building for agent payments, and put it live alongside x402 on one endpoint. This is what it actually looks like to take an agent's money over MPP today.

What we shipped

There is one endpoint, a payment front-door at agents.stacktr.ee, and it takes agent money three ways. An agent posts the HTML it wants hosted and gets back a 402 with the price, the assets, and the recipient. It can pay fifty cents over MPP on Tempo, or over x402 on Base or Solana, whichever it holds. On settlement the page is published to a private link and the response hands back the URL. No account, no card, no person. The endpoint is listed on mppscan and x402scan, the directories agents use to find paid resources, so an agent can discover it and pay it without ever being told it exists.

We proved it with real money on mainnet, not a sandbox. An agent paid fifty cents over x402 on Base, and the settlement is a USDC transfer on BaseScan you can verify. Another agent paid the same fifty cents over MPP on Tempo, and that settlement landed in our wallet the same way.

What MPP is, and why the onchain method is the interesting part

MPP, the Merchant Payments Protocol, is the standard Stripe and Tempo are building so software can pay a merchant per request. It answers a real gap in the agent economy: an agent has no card and no checkout to click, so it needs a machine-readable way to pay. MPP gives it one, and it supports more than one method underneath.

The method we run is the onchain one, and it is the part worth understanding, because it is permissionless. To accept an MPP payment on Tempo, a provider needs a receiving wallet, the public Tempo RPC, and a key it generates itself. That is the whole setup. There is no account to apply for and no regional approval to wait on, so a small company anywhere in the world can start taking agent payments over MPP the same afternoon it decides to. For anyone building on the merchant side of agent payments, that is the unlock: the onchain method turns accepting stablecoin from agents into a few lines of config rather than an onboarding process.

Gas is handled for the payer, too, so the agent only signs; it does not need to hold the chain's native token to pay. It signs an authorization, the settlement broadcasts, and the stablecoin moves. From the agent's point of view, paying is a single signed message.

Multi-rail: agents pay with whatever they hold

We did not pick one rail, because agents do not all hold the same thing. Some carry stablecoin on Base, some on Solana, some settle over Tempo. So the front-door advertises all three on one endpoint and lets the agent choose. It reads the 402, sees the options, and pays with whatever is in its wallet. From the agent's side it is one call; from ours it is three rails sharing a single publish flow. The aim is to never be the reason an agent cannot pay: if it holds stablecoin somewhere we accept, it can publish.

What is live

All three rails are live on the endpoint today: x402 on Base, x402 on Solana, and MPP on Tempo, settling real stablecoin on mainnet, with the receipts landing in our wallet. The endpoint is public and discoverable, so the fastest way to see it is to point an agent at it and watch it pay.

The broader bet is the same one as the first piece, now a rail further along. Your next customer may not be a person. It may be an agent with a wallet and a job to do, and MPP and x402 are how it pays you for the work. Being an early example of that, rather than waiting to see how it settles out, is the whole reason we shipped it.