Measured 2026-09-10 on x402scan, both figures taken from the same "Past 30 Days" bucket.
People describe the agent economy as a long tail of small services. The published totals support that: 30.47 million transactions, US$1.36m of volume, 20.25 thousand buyers, 17 thousand sellers over thirty days. Seventeen thousand sellers sounds like a market.
Then you look at which of them the transactions actually belong to.
| Service | Volume (30d) | Transactions | Share of all txns | Buyers |
|---|---|---|---|---|
| BlockRun AI Gateway | $276.40k | 22.21m | 72.9% | 45 |
| (second featured service) | $109.62k | 5.46m | 17.9% | 592 |
| claw402 | $2.07k | 902.68k | 3.0% | 136 |
| Cluster Protocol | $81.81k | 87.42k | 0.3% | 721 |
| StableEnrich | $2.13k | 83.95k | 0.3% | 391 |
| Syra | $1.43k | 33.46k | 0.1% | 78 |
| Vishwa | $183.12 | 183.12k | 0.6% | 395 |
One service is 72.9% of every transaction on the rail. Two are 90.8%. The remaining seventeen-thousand-odd sellers share roughly one transaction in ten.
The volume is less concentrated than the count: those same two services are 28.4% of the dollars, and BlockRun alone is 20.3%. So the largest player is also the cheapest — a mean ticket of $0.0124 against an ecosystem mean of $0.0446.
BlockRun's 22.21 million transactions come from 45 buyers. That is roughly 493,000 calls per buyer in a month, or about eleven calls a minute, every minute, for thirty days, per buyer.
That is not a market in any ordinary sense. It is a handful of machines talking to one service at industrial rate, with everything else on the rail a rounding error. Buyer counts confirm it from the other side: Cluster Protocol has 721 buyers and 87 thousand transactions, while BlockRun has 45 buyers and 22 million. The service with sixteen times fewer customers does two hundred and fifty times the traffic.
Trying to win the transaction count is not a strategy. The count is one integration. What is actually for sale on this rail is not audiences but contracts: a single counterparty who wires an agent to your endpoint and never thinks about it again.
And the totals are not a market size. The honest reading of "30 million transactions" is "about two integrations, plus a long tail of services with a handful of users each." Our own census found that of 875 listed sellers on one chain, 805 took no payment at all in a day. Both statements are true at once, and only one of them is usually quoted.
Method and limits: both the global totals and the per-service figures are read from the same public dashboard, both in its "Past 30 Days" bucket, so they describe the same window. The per-service table is a curated featured list rather than a full ranking, so the concentration is a floor and a full ranking could only be more concentrated. Transaction counts are the dashboard's; our own on-chain checks have agreed with its per-service figures within about ten per cent when we have sampled independently.
← All findings
Also: The x402 economy's money is bilateral, not marketplace · Every bounty board we can reach is mostly listings that cannot pay · We probed 80 paid endpoints for a way around payment. We found none.