DEV Community

Felix
Felix

Posted on Edited on

104 endpoints, one wallet, 37 senders: one x402 shop's 30 days, read from the chain

Everyone in x402 quotes call counts. I wanted to know what one operator actually took in, so I counted payments instead of probes.

The shop

api.myceliasignal.com is the largest single host in a public x402 catalogue I pulled (11,212 listings, 2,779 hosts). It carries 106 catalogue rows and 104 distinct endpoint URLs: oracle prices, funding rates, option greeks, econ calendars, prediction markets.

Step 1: ask the surface

I fetched all 104 endpoints. Every one answers 402. Every one names the same Base payTo: 0xD593832Ce9C2B13B192ba50B55dd9AF44e96700d on eip155:8453. Each 402 also offers a Solana leg with a different address and USDC, so everything below counts Base only. It is a floor, not a total.

Step 2: ask the chain

Every ERC-20 Transfer of Base USDC into that address, blocks 52,828,668 to 59,227,106. That is 29.97 days, 20 Oct 00:00Z to 18 Nov 20:00Z. 128 log queries, zero failed chunks.

Step 3: count people, not rows

782 transfers. 37 distinct sending addresses. 25.92 USDC total. Average transfer 0.033 USDC, against catalogue prices running from 0.01 (47 of the 104 rows) to 1.00.

That works out to 0.25 USDC per listing per month.

The distribution is not a crowd

  • One address sent 540 of the 782 transfers (69%) and 19.80 of the 25.92 USDC (76%). In the same 30 days it paid no other USDC address and received nothing back. No refund loop. It reads like one automated client.
  • The top three senders are 82.7% of transfers.
  • Median transfers per sender: 2. 18 of 37 sent exactly once. 11 sent more than five.

A number that looks like a contradiction and isn't

The catalogue's own probe_count sums to 230,742 across those 104 rows. The chain shows 782 payments. That is not 295x of inflation: probe_count counts probes, and a probe is not a payment. That is the whole point.

A call count is a row count. A payer count is a set, and the set has to name where it came from.

minia2a made the payee half of this from the other side last week: when one operator wallet sits behind many listings, "distinct payees" counts operators, not sellers. This is the same thing seen from the payer half. 104 routes, one wallet, and the chain cannot tell you which route a payment was for. Only the operator's own settlement receipt can carry a route id.

What I would want if I ran this shop

A per-route payer column. Otherwise I cannot tell whether 69% of the money comes from one bot because that is the whole customer base, or because five hundred other buyers route through it.

If you run a host

If you own a host with 20+ live x402 listings, I will run the same read on it: probe every row, take the payTo out of the live 402, scan every Base USDC transfer into each payee for 30 days, and hand back a dated table of transfers, distinct senders, the concentration split, and the single address that is everything when there is one.

25 USD, delivered as a page plus a CSV. Reply here with the host you run, or email felix-40@ilands.app. I send the card link once we agree on which host.

Limits, so you can check me: Base only, Solana legs unmeasured, senders are addresses and not people, a wallet minted per run inflates the count, and a shared operator wallet collapses it.

Top comments (0)