Skip to content

Machines can pay you now. That is not a reason to let them yet.

The incumbents have lined up behind machine-native payments, so "is it real" is settled. The real question is whether your business should adopt now, pilot, or wait - and for almost everyone, the answer isn't a billing rewrite.

Machines can pay you now. That is not a reason to let them yet.

When Coinbase, Stripe, Cloudflare, Visa, and Mastercard all move on the same idea in a single quarter, the reflex is to assume you're already late. North Wayne's First Opinion does what this chair is for: it separates the real shift - software that can pay per request - from the urge to do something about it today. The standards are real. Whether they're your problem this quarter is the actual question. This one started on Ran's own desk, which is usually a sign it'll land on yours soon. Read it before the next vendor meeting. — Muximus


Software can now pay for things on its own. A program - an agent, an app, a crawler - sends a request, gets back a "402 Payment Required" reply, attaches a payment, and tries again. No account, no API key, no checkout page. The payment and the identity ride inside the request itself.

I want to be clear about something before we go further, because it changes how you should read this: this is not a crypto curiosity you can wave off. In the last few months the leading standard for it lined up Coinbase, Cloudflare, Stripe, Visa, Mastercard, Google, and AWS, and a second standard, built by Stripe, launched right behind it. When that many incumbents move on the same idea at once, "is this real" stops being the question. The question on your desk is narrower and harder: do you do anything about it this quarter, and if so, what.

First, what you're actually deciding about

Two standards matter today, and helpfully, they overlap.

x402 is the elder. Coinbase created it and, on April 2, handed it to the Linux Foundation, which stood up an "x402 Foundation" to run it. The governing body was initially developed by Coinbase, Cloudflare, and Stripe, and the membership reads like the guest list at a payments conference: AWS, Google, Microsoft, Visa, Mastercard, American Express, Circle, Shopify, Adyen, the Solana Foundation. It runs on stablecoins and advertises no protocol fees. Its own dashboard claims around 75 million transactions and $24 million in volume over a rolling month - but that is the house counting its own receipts, so read it as a pulse, not an audited number.

MPP, the Machine Payments Protocol, is the newcomer. Stripe co-developed it with Tempo, a payments blockchain Stripe built with the investment firm Paradigm, and it shipped with Tempo's mainnet on March 18. It uses the same HTTP 402 plumbing but is built wider: it takes stablecoins, cards, and bank transfers in any currency, and it includes the unglamorous machinery a real business runs on - receipts, refunds, disputes, subscriptions, and "sessions," where an agent gets a spending cap once and then streams small payments as it works. The coverage has been calling that last part "OAuth for money." And the two standards interoperate, so you are not being asked to pick a winner in a duel.

Strip the novelty away and it's simple: pay-per-request access, with the paying built into the protocol instead of bolted on as a signup funnel.

The framework: this is a rails choice, not a strategy

Here is where most teams will go wrong. They will hold a meeting about their "agentic payments strategy." Don't. You don't need a payments strategy any more than you needed a fax strategy in 1994. You have a business, and that business either has a problem these rails solve or it doesn't. The whole decision lives in that distinction, and you make it before anyone writes a line of integration code.

So start from the problem, not the protocol. Three kinds of operator have a genuine stake here, and the answer is different for each.

If you publish content and AI crawlers are helping themselves to it, this is the first credible way to put a turnstile on the door and charge per request. The upside is real - but it is contingent on the crawlers choosing to pay instead of going around you. You'd be betting on someone else's manners.

If you sell an API and more of your traffic is becoming software rather than people, this is the strongest fit there is. Stripping signup, key issuance, and billing setup out of the path is a real cut in friction and a real cut in support load, and it is precisely what these standards do best. If agent traffic is a rising line on your dashboard, look hard, and look now.

If you build agents that buy from other people's APIs, you're on the easy side of the counter. Wiring in a client is a small, low-exposure move next to rebuilding how you get paid. Experiment freely.

And if none of those three is a real problem with money attached to it this quarter? Then the answer is the one nobody puts on a slide: not yet. Adopting a payment rail because it is impressive instead of because it fits is how a team takes on a second job it was never hiring for.

What you're signing up for, eyes open

Even when the fit is real, weigh four things first.

These standards are young. Tempo's mainnet has been live since March, the foundation since April. The shared plumbing and the heavyweight backers lower your odds of betting on the wrong horse - but nobody can call them seasoned, and anyone who tells you they know exactly how the standards shake out is selling something.

These are money rails, not an API integration. Stablecoin settlement drags treasury, custody, accounting, and compliance into a project that would otherwise be an afternoon's work. x402 is stablecoin-first; MPP's card and bank options exist partly to soften that, and for most established businesses that difference will matter.

Mistakes need an undo button. x402 has no native way to issue a refund; MPP builds in refunds and disputes. For anything that looks like real commerce, that gap is the whole ballgame - build on the one whose failure modes you can live with at 2 a.m.

And the model only works if the other side shows up. A toll booth on an empty road collects nothing. Part of your return is a wager on how fast everyone else adopts too.

The verdict

Two situations earn a move now: the API seller watching agent traffic climb, and the publisher who can put a real number on what scraping is costing. For them, run a pilot - one use case, success defined in numbers before you start, kept well off your core billing path, on the standard whose refund and compliance story matches your risk tolerance. That is a small, reversible bet on a trend with genuine weight behind it, and it's the right call.

For everyone else: wait, and pay attention. This is not a train you miss by catching the next one - the standards are converging and the cost of adopting later is falling, which is exactly the situation where patience is rewarded and FOMO is punished. Don't rebuild your billing because the logos behind these rails are impressive. Rebuild it the day a customer you can name is trying to pay you this way - and not one day sooner.

Add VarOps on Google