The ledger: what actually gets paid on x402

2026-08-30. For a month this directory showed each seller's on-chain activity as a number that never went above 150. It turned out to be the size of a third party's page, not a count. So we built our own index of every USDC Transfer on Base to every listed payment address, daily, and then did the thing no published “stablecoin inflow” figure does: told x402 payments apart from everything else arriving at the same addresses. This is what the chain says.

Thirty days, all listed sellers, Base USDC

7,497,574 payments. $670,111. 11,044 distinct paying wallets. 308 payment addresses received something. Window: 2026-07-31 to 2026-08-30 UTC (today partial), every listing with an EVM payTo in its 402 challenge.

Two economies share the rail, and a single sum hides both. Payments per day peaked at 1,228,585 on Aug 17 and fell to 13,136 by Aug 24 — a 99% drop, all of it one seller. Dollars per day stayed flat, $15k–$37k every day of the month, because dollars are carried by a handful of sellers taking $100 tickets while counts are carried by compute sellers at a cent or two a call. The median paid address received 67 payments and $0.95 in the month. Six addresses received more than 10,000 payments; one received 95% of all of them.

Which of it is x402

USDC arriving at an address is not x402 revenue. Some listed sellers also take ordinary payments at the same address. The chain can tell: the x402 exact scheme settles as an EIP-3009 transferWithAuthorization submitted by a facilitator, and that function selector is in every transaction. We sampled the 25 most recent payments to each of the 308 addresses (6,425 transactions) and classified each as x402-signed (EIP-3009), plain transfer (a wallet sending money the ordinary way), or other (any other calldata — a batching facilitator, a smart-wallet execute, a bridge; we don't classify it).

By count, the economy is x402: 7,479,218 of 7,497,687 payments (99.75%) are x402-signed. 108 addresses are pure x402 in the sample; 18 have none; 182 are mixed.

Update, 2026-08-31: the “other” bucket was largely our own blind spot. EIP-3009 defines each function twice — a (v,r,s) overload and a packed-bytes-signature overload — and our classifier only recognized the first. After adding the second (selector confirmed against USDC's verified contract source and computed independently), unclassified payments fall from 19% to 4% of the sample, and 95% of sampled payments are x402-signed. The remaining 4% is now mostly named rather than unknown: Multicall3 batch calls and ERC-4337 smart-account operations — wrappers whose inner calls we don't decode — plus one seller's unidentified custom contract. The dollar split barely moves (x402-signed ≈$247k of $671k): the misfiled traffic was micropayment volume. Found by decoding our own residue; the error direction was against x402, not for it.

By dollars, it isn't: about $250,000 of the $670,000 is x402-signed. About $392,000 arrives as plain transfers — nearly all of it at one gift-card merchant's payTo, where 22 of 25 sampled payments were ordinary checkout — and about $28,000 as other calldata. The dollar figures are sample-weighted estimates and are labelled as such on every listing; the counts are exact.

Named, because the chain already names them

blockrun.ai (10 listings, one payTo): 7,142,386 payments in the window, $166,968, 408 paying wallets, 25 of 25 sampled x402-signed. It ran at 100k–470k payments a day for three weeks, peaked at 1.2 million a day on Aug 17–18 — matching its own public “crossed a million a day” — and then fell 99% on Base from Aug 24 with its payTo unchanged. Base USDC only: whether that traffic moved chains or stopped is not something this index can say, and we don't. We showed it as 149.

api.bitrefill.com (6 listings): $435,419 over 3,792 payments from 2,709 wallets — the largest by dollars and by payer count, and mostly not x402-shaped (3 of 25). It is a gift-card merchant whose checkout address also answers 402s. Gross inflow, not x402 revenue; which is exactly the distinction this page exists to make.

api.nansen.ai (20 listings): 4,770 payments, $4,809, 181 wallets, 25 of 25 x402-signed. api.botpay.network: 24,549 payments from 3,804 wallets — the widest payer base on the network, at $1,688. stableenrich.dev: 58,245 payments, 429 wallets, $1,656. x402.twit.sh: 131,502 payments, 172 wallets, $1,077. laso.finance: $30,195 over 388 payments, 12 of 25 x402-signed. x402.clashofcoins.com: $19,494 over 830 payments, 0 of 25 x402-signed, all “other” — a contract we don't classify.

Every listing page now carries its address's daily series, its payment shape with the sample size, and how many listings share that address (figures belong to the address, not the route; never sum them across listings).

Against what's published

Every “stablecoin inflow to agents” figure we have seen is gross inflow to declared wallets, shape unknown, including ours until today. For comparison, one widely cited index reports $258,751 of inflow across six chains to 2,177 declared wallets over 30 days; on one chain, x402-signed only, listed sellers received about $250,000 — and 7.5 million payments against that index's 314,058. The difference is scope, not accounting: that index finds wallets through agent cards, and x402 sellers mostly don't have agent cards. Their number and ours measure different populations; ours also says which payments are x402.

Method, and what we can't see

Source: USDC Transfer events on Base whose recipient is a listed payTo, read from the chain directly and kept current continuously. Counts, distinct payers and dollars are exact over any window — there is no page limit and nothing is extrapolated; only the x402-signed / plain / other split is sampled, at 25 payments per address, and every figure derived from it is labelled as an estimate with its sample size. Payment shape is decided by the transaction's function selector: EIP-3009 transferWithAuthorization or receiveWithAuthorization is the x402 exact scheme's on-chain signature; transfer is an ordinary wallet payment; anything else is left unclassified rather than guessed.

Controls, because zeros and totals both need them: sampled transactions were checked against a public block explorer to the cent before any of this was published; the index reconciles to the third-party figures it replaced wherever that source was neither capped nor reading a different window, and reads higher exactly where it was capped. Anyone can check any figure here against the chain independently — the addresses are public, the window is stated, and the per-listing numbers are on each listing page and in the API.

Scope: listings whose 402 challenge declares mainnet Base (eip155:8453, or the spellings base and 8453 that some sellers use) — 1,489 of the 1,522 listings that declare a payment address at all. The other 33 (Solana 19, Polygon 8, Base Sepolia 3, Avalanche 2, X Layer 1) get no on-chain figures rather than figures from the wrong chain: three of them had payTos that receive real mainnet USDC while their challenge asks for testnet, which is a true number attached to the wrong claim. Removed 2026-08-30.

Not seen: any chain but Base; any asset but USDC; a seller whose payTo rotates per invoice (one listing rotates ~88 times a day and is excluded as ephemeral); x402 schemes that don't settle as EIP-3009. Shape is sampled, never enumerated. Shared addresses are attributed to the address. Dates of every correction to the numbers this page replaces are on /methodology (2026-08-29, entries 6–8).

Machine-readable twin: /state/the-ledger.json. Per-listing: onchain_daily, onchain_shape, onchain_payto_shared_listings on GET /v1/listings/:id. CC-BY-4.0; cite the URL.

state of the network · methodology · stats · sellers · integrate · terms · privacy