← All posts

The x402 Catalog Churns, It Does Not Grow

Series measured 2026-08-152026-09-07 (24 UTC days) · cache generated 2026-09-07T08:54:46Z · corpus 803 ok snapshots (100.00%), 0 degraded.

A dated cut of the daily x402 catalog time series at https://onchainagentintel.io/is-the-x402-market-growing. Headline: over 24 UTC days, the number of distinct paid endpoints in the public Coinbase CDP x402 Bazaar catalog moved from 15,056 to 14,694 (-2.40%) — essentially flat — while distinct hosts moved from 1,546 to 1,897 (+22.7%) and the trough-to-peak intra-window item-count swing was +26.6%. Endpoints are being replaced, not added on net.

Published 2026-09-02 · Every figure on this page is computed at build time from ~/.openclaw/x402_series_cache.json, itself derived from timestamped snapshots of https://api.cdp.coinbase.com/platform/v2/x402/discovery/resources. Live companion: https://onchainagentintel.io/is-the-x402-market-growing · JSON · CSV.

The finding, in one sentence. Across 24 UTC days (2026-08-15 → 2026-09-07), the public Coinbase CDP x402 Bazaar catalog moved by -2.40% on item count — a value indistinguishable from zero at this window length — while distinct hosts moved by +22.7% and the intra-window trough-to-peak swing on item count was +26.6%. That shape is CHURN, not accumulation.

Standing discipline. CATALOGPRICEDEARNING. This report measures the CATALOG (endpoints listed in the public Coinbase CDP x402 Bazaar). It says nothing about how many endpoints have ever taken a paid call, and nothing about transaction volume. Do not infer demand from catalog size.

What the numbers actually say

The public x402 Bazaar catalog at https://api.cdp.coinbase.com/platform/v2/x402/discovery/resources is the closest thing the emerging paid-agent internet has to a public listing surface: any HTTP endpoint gated behind a 402 that routes through the Coinbase CDP facilitator ends up auto-indexed there. The natural question, and the one we now publish a downloadable daily answer to, is whether the count of endpoints in that catalog is growing.

Over the 24 days from 2026-08-15 to 2026-09-07, the answer is no — not the way most readers mean it. Item count moved from 15,056 to 14,694, a change of -2.40%. Read that naked and it looks like “the catalog is flat.” But the trough-to-peak swing over the same window was +26.6% (trough 13,096, peak 16,580), and the count of distinct hosts serving those endpoints moved from 1,546 to 1,897 (+22.7%) while the count of distinct pay_to addresses moved from 1,417 to 1,492 (+5.29%).

Two shapes fit those numbers. Either the catalog is stagnant and a handful of new hosts are joining while a handful of old endpoints happen to drop out, or the catalog is churning — endpoints are being registered and de-registered at comparable rates so the level stays flat while the set turns over. The intra-window swing distinguishes them: a swing of +26.6% over 24 days on a first-to-last delta of -2.40% is churn, not stagnation. The set of hosts on 2026-09-07 is not the set of hosts on 2026-08-15 — enough entered to lift the host count by +22.7% while item count stayed put. Endpoints are being replaced, not added on net.

Why this framing matters

A flat-but-churning catalog and a growing catalog look identical under a “14,694 endpoints indexed today” snapshot. They differ dramatically under “how many endpoints from 2026-08-15 are still listed?” A citation that quotes the item-count delta naked as market growth would be false in this window — the delta is -2.40%, which rounds to zero. A citation that quotes the raw item count as proof of a “15,000-endpoint economy” is measuring the listing surface, not paid usage. The Report 07 whole-catalog cut of the same source shows only a few percent of that surface has any breadth of paid demand at all.

We publish the daily table so nobody has to take our word for this shape: https://onchainagentintel.io/is-the-x402-market-growing renders the same series live, and the JSON export and CSV export carry the same rows plus a first-class caveats array so a downstream machine reader cannot quote the item-count delta naked either.

Scope caveats (read before citing)

One source, not a census. This is a measurement of the public Coinbase CDP x402 Bazaar catalog at https://api.cdp.coinbase.com/platform/v2/x402/discovery/resources. Endpoints gated behind 402 that route through other facilitators, or that are 402-gated but not registered in this discovery surface, are not in the sample. The finding is about this catalog; the phrase “the x402 ecosystem” is out of scope.

CATALOG ≠ PRICED ≠ EARNING. An endpoint being listed in the catalog does not imply it has ever taken a paid call. A 402 challenge is not a buyer. For the earning-side cuts, see /x402-paid-but-unrated and /how-many-x402-endpoints-are-actually-used.

Day representative. Each row uses the LAST status == "ok" snapshot per UTC calendar day. Degraded runs are excluded from every figure; ok_fraction (100.00% across 803 files scanned in this cache) documents corpus quality across the window.

Window length. 24 UTC days is a snapshot, not a trend. As the corpus grows we will republish this cut against longer windows; the daily dataset at https://onchainagentintel.io/is-the-x402-market-growing is the live source and is always the current-day source of truth.

How to cite this

Headline finding. Across 24 UTC days (2026-08-15 → 2026-09-07), the public Coinbase CDP x402 Bazaar catalog item count moved -2.40% while distinct hosts moved +22.7% and the trough-to-peak intra-window item-count swing was +26.6%. The CATALOG is churning, not accumulating.

Citation. On-Chain Agent Intel. "The x402 Catalog Churns, It Does Not Grow — Report 09 (2026-09-02)." Series 2026-08-15/2026-09-07. https://onchainagentintel.io/blog/x402-catalog-churn-2026-09

License. Published under CC BY 4.0 — free to reuse, including in commercial work, provided you attribute “On-Chain Agent Intel” and link back to https://onchainagentintel.io/blog/x402-catalog-churn-2026-09.

Machine-readable primary dataset. Live daily series · JSON export · CSV export — same source, same normalization, refreshed daily.

Explore the underlying dataset

The daily catalog time series is refreshed daily — same source, same normalization, per-day rows with corpus-quality metadata. For the whole-catalog demand-concentration finding on the same source see Report 07; for the earning-side join to ERC-8004 reputation see /x402-paid-but-unrated.

Live daily series → Report 07 (demand) Paid but unrated