x402 on Base: a $30 median, or 3 cents?
On 5 November a Coinbase Developer Platform engineer posted that x402 payment volume on Base is going "up, to the right", and that the payments are getting bigger too, with a $30 median: real businesses whose users expect their agents to buy on their own. We count x402 payments on Base every hour as well, for the payout addresses in the x402 Trust Index. Those are the sellers listed in public catalogs, the open side of x402 that any agent can find and pay. On that side, volume is rising fast too. The payments, though, are tiny.
Up and to the right: yes
In our first full count, 20 to 22 October, publicly listed sellers on Base received about 48,800 payments and $2,280 a day (x402 runs on cents, but the money comes from a few big payments). From 30 October to 5 November it was about 174,000 payments and $4,920 a day: 3.6 times the payments and 2.2 times the money in two weeks. Part of that is the catalogs growing: we watch more payout addresses now than in October.
Bigger payments: not on the open side
Payments grew faster than money, so the average payment fell from 4.7 cents to 2.8 cents. This is where the payments went, grouped by what each seller averaged per payment that day:
| Seller's average payment | Payments | Share | USD | Share |
|---|---|---|---|---|
| Under $0.10 | 1,204,085 | 98.9% | $19,886 | 57.8% |
| $0.10 to $1 | 8,176 | 0.7% | $1,424 | 4.1% |
| $1 to $10 | 4,913 | 0.4% | $10,807 | 31.4% |
| $10 and more | 64 | 0.005% | $2,298 | 6.7% |
The typical payment, the median, is about two cents. Only 10 sellers averaged $10 or more on any day, and only 4 averaged $30 or more, with 7 payments between them. The bigger payments do bring in real money: the 54 sellers averaging a dollar or more had 0.4% of the payments and 38% of the dollars.
The open side is also very concentrated. One payout address received 58% of all payments in the week, and three received 84%.
Not every payment is a purchase
- Checkers and scanners. 52 wallets paid ten or more different sellers in the week: monitoring tools, catalog crawlers that test-buy, harnesses. 985 of the 1,655 sellers that were paid at all got money only from wallets like these (we wrote about this in Half the x402 sellers paid on Base were paid only by scanners).
- Money going round. 53 sellers sent at least half of what came in back to the wallets that paid them, a closed loop that looks like a busy seller on chain. In all, $4,642 went back to 270 of the 7,910 paying wallets.
Why a $30 median and 3 cents can both be true
We count different sellers. Public catalogs, such as the Bazaar, list open pay-per-call APIs: a price, a search, a model call, for a cent or two. Businesses that use x402 for larger purchases, a report, a booking, a batch of work, often aren't in any catalog: their agents and their customers already know where to pay. Coinbase's facilitator settles payments for many sellers on both sides, so it sees what we can't. A median also ignores how many tiny calls there are: a hundred thousand two-cent calls a day to one seller don't move it, but they decide a payment count.
So the open market is still mostly agents paying cents for data. The larger payments are growing somewhere else, mostly out of public view.
Why bigger payments make a check worth more
At two cents a call, a wrong payment costs two cents, and few people bother to check the seller first. At $30 it's different. A payout address that changed overnight, a price several times what the seller advertises, a seller that takes the money and doesn't answer: each of these now costs real money. In our own test buys from random sellers that a check passed, 74 to 89% delivered (Random x402 sellers delivered 74 to 89% of the time).
What an agent should check before a large x402 payment:
- The seller's payout address: is it the one it was yesterday, and is it bound to the seller's identity (an ERC-8004 agent wallet or a payee delegation)?
- The price: does the 402 ask what the seller advertises, and what it asked before?
- The track record: has this payout address been paid by ordinary buyers, or only by scanners and its own wallets?
- The payment itself: amount, token, network and validity against the 402, before signing.
A x402 Doctor preflight does the first three for $0.001, which is 0.003% of a $30 payment. presign-guard checks the payment itself before signing. The fraud-classes guide says which kinds of payment fraud a check before paying catches and which it can't.
How we counted
- x402 Doctor reads every Base block for USDC transfers to the payout addresses in the x402 Trust Index (3,402 on Base this week). Only transfers the payer signed with EIP-3009 (
transferWithAuthorization, the way x402's exact scheme settles) count. Plain transfers, Permit2 and Circle Gateway batches don't. x402 Doctor's own test wallet is left out. - The window is 30 October to 5 November 2026, UTC. The count is open data: payments-data on GitHub, and daily on Hugging Face.
- We store a count and a total per payout address per day, not each payment. So the size bands use each seller's daily average, and the median of about two cents is the median of those averages, weighted by payments.
- Wallets that paid ten or more different sellers in the window count as checkers and scanners. A closed loop is a seller whose ordinary payers all got USDC back from it, at least half of what came in.
- Base went from 2-second to 0.2-second blocks on 5 November around 17:55 UTC. Our counter assumed 2 seconds, which put some evening payments on the wrong hour, but within the same week. It now measures the block time (x402-doctor#540). The 7-day totals are not affected.
- We don't see payments to sellers that aren't in a public catalog. The Coinbase figure is theirs, and we can't check it.
Questions
How big is an x402 payment on Base?
It depends on which sellers you count. To the 3,402 payout addresses in public x402 catalogs, 1,217,238 signed USDC payments worth $34,414 were made from 30 October to 5 November 2026: 2.8 cents on average, and 99% went to sellers that averaged under ten cents a payment. A Coinbase engineer, who sees the payments Coinbase's facilitator settles, reported a $30 median on 5 November.
Why do the numbers differ so much?
They count different sellers. Public catalogs list open pay-per-call APIs, which mostly charge a cent or two. Businesses that use x402 for larger purchases often aren't listed anywhere, but many of their payments go through Coinbase's facilitator. A median also ignores the millions of tiny calls that dominate a payment count.
Are all these payments real purchases?
Not all. 52 wallets paid ten or more different sellers in the week (checkers, scanners, test harnesses), and 985 of the 1,655 sellers that were paid got money only from such wallets. 53 sellers sent at least half of what came in back to the wallets that paid them (a closed loop), $4,642 in all.
Check a seller before paying with x402 Doctor, check a payment before signing with presign-guard, and let an agent buy within your limits with the Fizzl Agent Wallet.