Where our numbers come from

402index.io — A directory of paid endpoints across L402, x402 and MPP — self-described as 15,000+, which would make it the largest single frame available to us.

Full enumeration is paid (L402-gated CSV export); the free API rate-limits into 402s well before its rows are exhausted. A partial read is unread by the population layer's own law, so it is named instead of half-read.

What would dissolve it: The wallet law was ruled 2026-08-18 ($25/month funding discipline). What remains is a funded wallet, one hand-captured paid response to build the reader against, and a price the keeper has seen — none of the three is done, and the price is not published anywhere we could reach.

x402scan.com — A settlement indexer with a resource directory; the only one of these that watches money move rather than doors exist.

Resource enumeration is a paid endpoint. Paying is a wallet decision.

What would dissolve it: The wallet law was ruled 2026-08-18; a funded wallet and one hand-captured paid response remain. Partial relief already landed: x402-list carries rows it imported from x402scan, tagged `imported:x402scan`, so a slice of this population now arrives free — 3 rows of 630 on 2026-09-04, which is a slice and not a substitute.

endpoint.x402jp.com — A crawler-built index of x402 hosts ranked by catalogued paid routes — 19,366 routes across 1,031 hosts when it was read by hand on 2026-09-03.

No response shape has been captured. The site is unreachable from every sandbox this store is built in, so the 2026-09-03 read-off in research/x402-pulse.md came from a pasted copy of the page and cannot be refreshed from a build.

What would dissolve it: One hand-run read from the keeper's browser, saved as a fixture, exactly like every feed before it. The store's own row there disagrees with the counter (61 routes against 39), so the capture is worth having for its own sake.

agent-tools.cloud — An aggregator of x402scan, awesome-x402 and the CDP Bazaar with self-submissions on top; 20k+ entries, liveness-probed, refreshed every six hours.

No response shape has been captured and no public enumeration endpoint is documented anywhere we could reach.

What would dissolve it: One hand-run read. Note before spending effort: it aggregates two frames we already hold, so its marginal population may be small — worth measuring against the union before it is built.

x402scout.com — A self-described canonical registry with 0-100 trust scores, MCP-native, rescanning every six hours.

No response shape has been captured, and the host is unreachable from every sandbox this store is built in. Note also what would NOT be read if it were: its 0-100 trust scores are a standing verdict on operators, and this store does not republish other people's rankings — only the host list would enter the population.

What would dissolve it: One hand-run read of whatever list endpoint it serves, saved as a fixture, with the scoring fields dropped at the parse rather than carried and ignored.

nohumans.directory — A directory that probes every listing every fifteen minutes and reports on-chain-proven buyers.

No response shape has been captured, and the host is unreachable from every sandbox this store is built in. It is the highest-cadence prober on this roster — every fifteen minutes against every listing — so its host list is the freshest of the free frames and the one most likely to disagree with our weekly picture.

What would dissolve it: One hand-run read. Its own prober already visits us, so the relationship exists in one direction and has simply never been read in the other.