How big is the x402 economy? We measured it

The number

Across the entire public x402 resource catalog on 7 September 2026: $7,630 per month of claimed payment volume, spread over 14,118 active listings and 1,073 merchants. The median listing reports two calls and about two cents a month.

That is the whole protocol, not a segment of it. Anything priced in x402 today is competing for a pool roughly the size of one enterprise software seat.

Why it is not the number you will see quoted

Read the catalog naively and it reports $37,916 — five times larger. The difference is one field.

Every catalog row sets its price and its ceiling to the same value, so an invoice maximum is indistinguishable from a per-call price. A single gift-card endpoint listed at $1,000 per call, with 28 recorded calls, carries about 74% of the unfiltered total on its own. It is almost certainly a ceiling on a variable-value purchase rather than a thousand dollars charged twenty-eight times. Excluding listings above $10 per call is the entire distance between $37,916 and $7,630.

We publish both figures because the correction is a judgement call and you may want to make a different one.

Concentration

• The top 8 merchants account for 53.3% of measured volume. The remaining ~1,065 split the other half.
• 74% of listings report exactly one unique payer — the signature of a merchant testing their own endpoint, not of a customer.
• Roughly 4% of listings account for essentially all repeat payment.

A long tail of listings with one payer each is a catalog of experiments, not a marketplace. We are in that tail ourselves.

Do the endpoints actually work?

Listing something is not serving it, so we probed every listing priced at $0.50 or above — 465 of them:

• 90.5% return a payable 402: a real challenge an agent could sign against.
• 4.9% are uncallable path templates — the listing advertises a URL with an unsubstituted placeholder in it, so every probe hits a URL that cannot succeed.
• 4.5% answer with a 4xx or 5xx that is not a payment challenge.

The infrastructure is in better shape than the demand. That is the opposite of the usual early-protocol failure, and it is worth saying plainly: x402 endpoints mostly work. Almost nobody is paying them.

The mistake we made first, in case you are counting too

Our first pass concluded that a fifth of the catalog returned a 402 no agent could pay, and that the ecosystem was largely broken. That was wrong, and the cause was a case-sensitive header lookup.

Most x402 servers send the challenge in a lowercase payment-required header alongside an empty {} body. Match the header name case-sensitively and those endpoints look like they are answering 402 with no terms attached — indistinguishable from genuinely broken. Read it case-insensitively and they are fine. HTTP header names are case-insensitive by specification; a census that forgets it will libel about 19% of the catalog. Ours did, for one round.

What it is safe to conclude

Two limits travel with these figures, and they are in the script's output rather than buried here. Catalog usage telemetry is self-reported by facilitators and never audited, so claimed calls are claims — and the single-payer concentration is a reason to distrust them upward, not downward. And a payable 402 proves an endpoint answers correctly today, not that anyone has ever paid it.

What survives both caveats: the ceiling on any x402-priced business today is small, known, and concentrated. That is a useful thing to know before building one, which is why it is published rather than kept.

Reproduce it

The script is in our repository, takes about two minutes, needs no API key, and has an offline self-check covering both traps above:

node scripts/x402-census.mjs             # catalog census
node scripts/x402-census.mjs --probe     # + liveness of the >=$0.50 band
node scripts/x402-census.mjs --selftest  # offline, no network

It reads the public CDP resource catalog. If your numbers differ from ours, the catalog moved or you drew the ceiling somewhere else — both are worth knowing.

Why we published this

We sell an x402-priced API. This measurement caps our own market and says so. We would rather be the source people cite for the size of x402 than be quietly wrong about it in our own planning — and a number that only flatters the person publishing it is not worth reading.

If you want to see the protocol work end to end at two cents a call, that is how the payment flow works and this is the service:

Try it without an account

curl -X POST https://slidingbox.ai/v1/key

Returns an evaluation key immediately — no signup, no email, no approval step. Send it as X-API-Key on a retrieval to skip payment while the key has allowance left. Storing is free either way.