Every other page here counts payments. This one counts the payers, and the answer is not what the volume figures imply.
1,360 addresses make 93.7% of every agent payment we read. There are 70,419 addresses that have ever paid.
Of those 1,360, 6 appears in an ERC-8004 registry and none is a vault we read. They paid 2,523 different counterparties, and the middle payment among them was $0.0100.
The bands are CHOSEN, not found.
This distribution has no natural breaks in it, so any edge is a decision; these are round numbers at the orders of magnitude the data spans. The counts are published per band so a reader who would draw the lines elsewhere can.
| payments each | addresses | payments | share of activity | value moved | |
|---|---|---|---|---|---|
| more than 1,000 | 1,360 | 17,225,335 | 93.7% | $282.9k | |
| 51 to 1,000 | 3,582 | 934,204 | 5.1% | $137.3k | |
| 6 to 50 | 7,628 | 113,234 | 0.6% | $21.3k | |
| 2 to 5 | 15,588 | 66,779 | 0.4% | $954.2k | |
| 1 payment | 42,261 | 42,261 | 0.2% | $5.28M |
The busiest band, more than 1,000, is 1,360 addresses moving $282.9k. The band holding the most money, 1 payment, is 42,261 addresses moving $5.28M.
So a figure for “the agent economy” counted in events and the same figure counted in dollars describe different sets of addresses. Neither is wrong, and a page that gives one without saying which is misleading by omission.
They send 5171 times more than they receive, to 2,523 different counterparties. That is the shape of a buyer, not of a market maker or a relay, and it is the clearest thing we can say about addresses that identify themselves nowhere.