stelar digital · census

We re-counted the x402 catalog today. It's smaller than it was in July.

Part of our x402 series: guide · pricing · vs RapidAPI · start here · quickstart · state of x402

Updated 2026-09-27, re-reading the catalog and our own logs the same day.

x402 is the payment protocol that lets an AI agent (software, often built on a large language model, that carries out a task on its own) pay for a call to an API (a web address one program calls to get data or trigger an action from another) with a stablecoin, per request, no accounts — explained in plain terms on the x402 guide. Coinbase runs a public discovery catalog of every registered x402 endpoint; independent trackers (Chainalysis, KPMG) have separately published their own volume analyses. This page is honest about which numbers below we re-read today and which are dated findings from July that we haven't re-checked.

What we measured (method first)

We queried Coinbase's public x402 discovery catalog (the same index agent buyers search) again today. The total below comes straight from the catalog's own pagination metadata — anyone can reproduce this, it's a public API. We did not re-run a full price-distribution resample today, so where we'd need one to back up a specific number, we've described the shape of the finding in words instead.

The supply side, today

• The catalog counted more than 17,000 registered endpoints today — down from the 25,443 we counted in July, so the catalog's makeup keeps shifting.
• In July, the endpoints we sampled from it skewed heavily toward very small per-call prices, rather than anything close to a typical human-facing subscription price. We haven't re-sampled the full price distribution today, only the total count.

The demand side (July findings, not re-read today)

As cited in our July 11 census, Chainalysis reported that on-chain x402 transaction value had shifted toward much larger individual payments over that period. Separately, KPMG's own agentic-payments tracker, published around the same time, counted far more buyers than sellers and a modest average lifetime settlement per seller — a different measurement than Chainalysis's, and we're careful not to blur the two together. We haven't re-read either tracker's current output today, so we're dating both as findings from our July 11 census — see their own published work for whatever they show now.

Put simply, as of July: a lot of supply looked priced well below where the reported demand was clearing. Whether that gap is still real today is exactly the kind of claim we won't restate without re-reading it.

Our own receipts (we're not spectators)

We've run priced endpoints with full catalog presence since early July, and we're honest about the small scale: as of 2026-10-17 05:30 UTC, real settled calls and real on-chain settlement, updated automatically. Live figures: /open.

We also tested the "premium bundle" theory ourselves — two combined-signal bundles priced well above our raw, single-purpose endpoints. We set a public judge date in July on whether that would catch on. It mostly hasn't: as of today the two bundles account for only a couple of our settled calls between them, out of dozens on our primary rail. The atomic, cheap fact is still what sells. See the full mechanics at how pricing works.

A second settlement rail. The same top data endpoints are now also live on a separate host on Algorand mainnet, settled via the GoPlausible x402 facilitator, at api-algo.stelardigital.com — a stablecoin per call, no accounts. It's live and serving signed 402 payment challenges now; the first on-chain settle landed 2026-07-23T23:37:32Z (tx DX6P544HF6T6…).

What today's re-count means

1. The rail works; the market hasn't finished forming. A large, actively-changing catalog is infrastructure proof, not demand proof — most registered endpoints have likely never seen a paying agent.
2. The catalog shrank, not grew, since July. That's the one supply-side fact we can point to from today's own read; we're not extending it into a pricing verdict without re-reading the demand side too.
3. The opportunity isn't panning — it's picks, shovels, and maps. With supply this large and, per KPMG's July count, far more buyers than sellers, seller-side tooling — delivery receipts, buyer analytics, honest catalogs — is the underbuilt layer. We maintain x402-receipts, an open-source, Apache-2.0 verifiable-receipts library, proposed as a protocol extension.

The franchise promise

We'll keep re-reading this census and publish the deltas: endpoint counts, our own settlement mix, wins or near-misses. If you sell on x402 and want seller-side tooling (know who's buying, prove what you delivered), email stelardigital@stelardigital.com with subject "seller console" — if enough sellers raise a hand, we build it; if not, we'll say that too.

Sources: Coinbase CDP x402 discovery catalog (re-queried 2026-09-27, 17,533 registered endpoints; queried 2026-07-11, 25,443). Chainalysis x402 on-chain analysis (transaction-value concentration finding) and KPMG's agentic-payments tracker (buyer/seller/revenue counts) — both their own published work as cited in our July 11 census, cited directionally and dated, not re-read today for exact figures.

Not financial advice. Any trading-related figures on this page reflect our own retired account. See our Disclaimer.