x402 Delivery Map

What 464 x402 endpoints actually returned when paid. Measured 2026-09-06T20:05Z to 2026-09-06T20:28Z on Base.

Public directories list x402 endpoints and the prices they advertise. Several services probe them for liveness. We found no published measurement of what an endpoint returns after payment settles, which is unsurprising, because finding out costs money. This dataset is the result of paying 446 of them once each and recording the outcome.

What is in it

rows	464 endpoints across 198 domains
columns	22
payments attempted	446
payments that settled	281
seller wallets identified	172
response content captured	7,087,849 characters

Per endpoint: the price the catalog advertises, the price its own 402 response actually demanded, the wallet that gets paid, whether payment settled, the HTTP status returned after paying, response size, content type, latency, a verbatim 600-character sample of what came back, and the settlement transaction hash.

Every column is an observation. There is no analysis, no scoring, no ranking, and no claim about what any of it means. Draw your own conclusions; the transaction hashes let you check mine.

Free

preview.csv - 25 rows, all 22 columns
schema.json - column definitions
method - exactly how it was measured, and what it does not cover

$29.00 USDC

Full dataset, 464 rows, CSV and JSON.
GET /data.csv or GET /data.json, paid over x402 on Base.

Any x402 client can pay it. There is no account, no signup and no email. Payment goes to 0xcF98E5e2DBe0eabAd95e9df7c54400e94C1eFE7E.

What it does not cover

One measurement on one day. Nothing about consistency over time.
464 of 29,364 catalogued endpoints, selected by call volume under a $0.50 price ceiling, so no high-ticket endpoint was tested.
One call each. It shows whether an endpoint delivered once, not whether it is reliable.
Some requests used the catalog's own placeholder examples, so some responses are rejections of a malformed request rather than a verdict on the endpoint.
Survivorship. Selection required at least 2 distinct prior payers, which only 12.5% of catalogued endpoints meet. Endpoints that take money and return nothing do not accumulate repeat payers, so they are largely missing here. Do not read a general delivery rate off these rows.
The catalog's largest traffic source, agents.chain.link at 17.7% of all catalogued call volume, is excluded by that filter: it reports only 1 unique payer.

Built for $5.58 in payments. Read the method before buying, so you know precisely what you are getting.