The Live x402 Population, Measured Not Estimated

Three denominators, kept strictly separate

The single-most common category error in circulating x402 commentary is treating the catalog, the priced set, and the earning set as the same number. They are not.

Same snapshot (2026-09-07T09:00:55Z). Read plainly: an endpoint appearing in the CDP Bazaar is not a paid endpoint, and an endpoint quoting a USDC price is not one that has been paid.

The price census refutes “~$0.30 average”

We publish the only whole-catalog census of x402 endpoint prices at /what-does-an-x402-api-call-cost. The population is heavy-left, not centered around a third of a dollar: median $0.0100, p90 $0.1000, p99 $1.25 — the “~$0.30 average” sometimes cited is arithmetic mean territory that no priced endpoint on the current snapshot sits near.

Denominator: 14,406 priced endpoints (of 14,689 indexed). 283 endpoints are excluded — they have no resolvable numeric USDC leg on this snapshot.

Payer- and call-weighted concentration, side by side

Report 07 published the endpoint-weighted view. Report 10 published the call-weighted view. The two together are the shape that matters: a small cohort of endpoints carries most of the requests, and the exactly-one-payer cohort — the visual majority — carries near-nothing.

Endpoint-weighted denominator: endpoints with BOTH a payer count AND a call count present. Call-weighted denominator: total 30-day indexed requests. Both denominators next to their numerators; the two answer different questions and cannot be conflated. A 1-payer endpoint is indistinguishable from self-dealing from on-chain data alone — not proof of it, and not proof of a real external customer.

ERC-8004 ownership — endpoint AND publisher level, together

This is the pointy question the CDP catalog cannot answer on its own: who owns the endpoints that are earning? We published the join finding at /x402-paid-but-unrated (measured 2026-09-01, snapshot 2026-09-01T11:23:10Z, chain base, sample defined below). The endpoint-level and publisher-level cuts belong in the same block — quoting one without the other is where the number becomes a talking point rather than a finding.

Sample: 14,338 x402 endpoints on Base with quality.l30DaysUniquePayers >= 1 in the 2026-09-01T11:23:10Z bazaar catalog snapshot. Every share is a lower bound — pay_to may be a treasury wallet distinct from the ERC-8004 registration owner; per-publisher dedup capped at 50 for the base-resolvable count. Canonical predicate: smartcontractauditpro/commerce_backed.py. Full derivation in /x402-paid-but-unrated.

The commerce-backed 0. The 0 endpoints and 0 distinct publishers on the commerce-backed line are not a data bug and not an argument that commerce-backed reputation does not exist. We index 44 commerce-backed agents — every one of them ships over Virtuals ACP, not as an x402 HTTP endpoint in the CDP Bazaar. The two surfaces do not overlap on today's deployed state. See the live commerce-backed cohort dataset for the got-paid cut in its own frame.

The catalog churns — it does not accumulate

A distinct claim from concentration or size: measured across 24 UTC days (2026-08-15 to 2026-09-07), the CDP Bazaar's item count moved by -2.40% while distinct hosts moved +22.7% and the trough-to-peak intra-window swing on the item count was +26.6%. That shape is CHURN, not accumulation — endpoints are being replaced, not added on net. See Report 09 for the long-form treatment and the same daily series behind /is-the-x402-market-growing.

What the population does not carry

Three signals that a critic would expect a mature x402 ecosystem to publish are missing on the population, not on the individual best-of-catalog endpoints. Naming what is not there is part of the honest measurement.

• Reputation. The ERC-8004 ReputationRegistry exists, feedback rows land, and independent measurement (Imperial College, arXiv:2606.26028) reports the registry cannot function as a trust signal in its current deployed state. On the base-resolvable join above, only 9 earning endpoints (across 4 distinct publishers) carry any registry feedback at all.
• Commerce-backing. Zero on the join, and the reason is co-located above — Virtuals ACP agents do not surface as x402 HTTP endpoints in the CDP catalog. This is a boundary condition of the two surfaces, not a rating of the cohort.
• Endpoint identity. The CDP Bazaar records a payTo address per accepts leg but not an owner identity that walks back to an on-chain registration. Our signpost-gap measurement is the walk from the identity side: 6.5% of indexed Base + Ethereum ERC-8004 registrations publish a discoverable A2A agent-card. The population does not, in general, carry a walkable identity.

Methodology and scope

Bazaar source. https://api.cdp.coinbase.com/platform/v2/x402/discovery/resources — the public Coinbase CDP x402 Bazaar discovery API. Unauthenticated, paginated JSON. Every entry carries an accepts[] array (asset + network + payTo + scheme) and a quality sub-object with l30DaysUniquePayers and l30DaysTotalCalls.

Snapshot. 2026-09-07T09:00:55Z. CDP’s reported catalog total on the same page fetch: 14,689. The catalog is fluid across snapshots; the shape (a large listing surface, a small paid-usage cohort, a heavy-left price distribution) is stable across recent snapshots.

Denominators. Every percentage on this page renders next to the two raw numbers that produced it. Endpoint shares use endpoints-with-both-fields; call shares use total 30-day requests indexed. Price percentiles use the priced-endpoint denominator, called out separately from the catalog total. The join figures use their own base-resolvable denominator, dated 2026-09-01, and every share is a lower bound for the reasons above.

What this is not. Not a census of every x402 endpoint on the internet — the CDP Bazaar is one Base-weighted discovery source and endpoints gated by 402 that route through other facilitators or are not registered here are out of scope. Not a settled-USDC figure — nothing on this page is “volume” in a dollar sense. Not a claim about wallet-level cumulative x402 activity across Base — that is a different unit of count on a different time window; the wallet-level cumulative view is separately reported by Chainalysis as an on-record third-party read and is directionally consistent with the concentration shape here without measuring the same object.

How to cite this

Headline finding. On the CDP Bazaar snapshot dated 2026-09-07T09:00:55Z, 14,689 endpoints are indexed; the priced subset has a median call cost of $0.0100 and a p90 of $0.1000; the ≥5-payer cohort is 5.19% of the with-both-fields set but carries 55.1% of indexed 30-day requests; on the base-resolvable owner join 26 distinct pay_tos matched, of which 4 carry any ERC-8004 feedback and 0 are commerce-backed.

Citation. On-Chain Agent Intel. "The Live x402 Population, Measured Not Estimated — Report 11 (2026-09-03)." Snapshot 2026-09-07T09:00:55Z. https://onchainagentintel.io/blog/x402-live-population-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-live-population-2026-09.

Machine-readable primary datasets. Live market map · JSON · CSV · price census (JSON, CSV) · paid-but-unrated join (JSON, CSV) · daily catalog time series. Same source snapshot as this piece unless the dataset carries a different fetched_at.