Research · Part II
It’s not one rail — it’s a dozen.
A month ago we published the first independent measurement of the x402 economy — agents paying for APIs in stablecoins — and were upfront about its one real limitation: it looked at a single rail, Coinbase’s x402 on Base. A neutral measurer that only sees one rail isn’t really neutral; it’s just that rail’s tool with better manners. So we taught the measurement to read a second standard. What it found is that the “which rail wins” framing everyone argues about is the wrong question.
The Machine Payments Protocol (MPP), co-authored by Stripe and Tempo, turns out to use the same core mechanic as x402: an unpaid request comes back as 402 Payment Required with the price attached. The difference is only where — x402 puts the challenge in the response body; MPP puts it in a WWW-Authenticate: Payment header that decodes to the amount, the token, the chain, and who to pay. Same idea, different envelope.
So we extended our probe to read both, and pointed it at the MPP registry too. Every number below now comes from measuring both stacks — Coinbase’s and Stripe/Tempo’s — the same way, on the same schedule, with the same public, Bitcoin-anchored record.
Here’s the number that reframed it for us. Of 629 services we can currently verify as alive and honestly priced, the settlement isn’t split between two rails — it’s spread across roughly eight chains:
A service listed under one protocol may settle on a chain associated with neither. Any picture of this economy that stops at one rail — or even two — misses most of the map; closing that gap is exactly why a neutral, cross-rail measurer exists.
Extending to MPP didn’t just add coverage; it added an honest contrast. Of the ~100 services in the MPP registry, 25 verify as alive and priced — a 25% rate, against 63% for x402. Nineteen of them settle on Tempo, Stripe’s new stablecoin chain, which went to mainnet only in March.
That gap is a finding, not a footnote. A newer registry carries more aspirational listings — services announced before they’re live — which is normal for a young standard. Part of it is also our own limitation, stated plainly: a chunk of MPP endpoints only respond to POST requests, and our unpaid probe currently uses GET, so we can’t yet trigger their challenge. We’d rather tell you that than quietly count them as broken. (Fixing it is on the list, and it lifts x402’s numbers too.)
Our headline “verified” count rose from 569 to 629 over the month. Some of that is real movement in the ecosystem — but part of it is simply that we started measuring a rail we couldn’t see before. We won’t dress up “we improved our own coverage” as “the market grew.” The same discipline held last time, when a parser fix — not the ecosystem — moved our numbers. A measurement is only worth what its willingness to separate the two is worth.
Behind all of this is the part that doesn’t make headlines: the machine has now run 73 sweeps over four weeks, each one git-committed and Bitcoin-anchored, an unbroken and tamper-evident history that no snapshot can fake and no newcomer can backfill. Reading a second rail made the picture wider; four weeks of continuity is what makes it trustworthy.
As agents start to move real money on their own, “which service should I pay, on which chain, at what real price?” becomes a question no single rail can answer neutrally — each has a reason to make its own ecosystem look busiest and healthiest. Independent measurement is only useful if it spans all of them and belongs to none. When your agent pays on your behalf, you’re accountable for it, written rules or not — and being accountable means being able to point to an independent record that the counterparty was real, fairly priced, and actually used, whatever rail it lived on. That record now spans the whole sprawl, not one corner of it.
Verantis is a neutral, cross-rail measurement layer for machine payments. Every sweep is public and anchored, and the data is open for you to check and poke.
Numbers are from the Verantis sweep of 2026-08-05, spanning x402 and MPP services across all observed settlement chains. Methodology and the full sweep history are public and reproducible. Figures describe the ecosystem at that date; the live directory reflects current numbers.
This report presents independent measurements and our own analysis of them. Measured figures are point-in-time observations and may change; interpretations and any forward-looking views are our own and may differ from others’ assessment of the market. Nothing here names or alleges wrongdoing by any specific service, and nothing here is financial, investment, or legal advice.