We measured settled volume on-chain across five networks. Of ~$50,000 in gross agent-to-agent payments, about 69% is single-payer self-funding or test traffic. Here's the receipt.
Every few weeks a new headline puts a number on the "AI agent economy" — $50 million, $600 million, pick your press release. We wanted the number you can actually check: not what a marketplace reports about itself, but stablecoin that genuinely moved on-chain, into the wallets agent services get paid at.
So we measured it directly. Across Base, Polygon, Arbitrum, Ethereum, and Solana, we sum every USDC (and USDT/PYUSD) transfer into the payTo wallets in our catalog, attribute each payment to its sender, and record it forward, every day. No estimates, no self-reported volume.
Here's what the raw number looks like right now:
| Measure | Amount |
|---|---|
| Raw settled volume (all chains) | $49,971 |
| Single-payer self-funding + non-production | $34,524 (~69%) |
| Validated settled volume | $15,447 |
The headline figure is real on-chain data. It is also mostly wash.
The single largest recipient of settled volume on Solana is a wallet attached to a service literally named ponzi402.com. It took in $30,730 — of which $30,245 came from one payer, in one transaction.
That is not an economy. That is one wallet paying itself. A real service has many payers; a wash wallet has one. When 98% of a wallet's inflow comes from a single counterparty on five figures of volume, it isn't demand — it's a loop.
Right behind it: a service whose endpoint is localhost:3000 — a developer's laptop — booking nearly $1,900 in "settled volume."
Strip out the single-payer self-funding and the non-production endpoints, and the validated floor is about $15,447. Most of that is one genuine business — Bitrefill, a real merchant selling gift cards and top-ups — earning steadily across multiple chains from many different payers. That's what real demand looks like: distributed, repeated, external.
So the honest one-liner: the measurable agent economy today is a low five figures of real settled volume, concentrated in a handful of genuine services, wrapped in a much larger cloud of self-dealing and testing.
We don't hide the raw figure and we don't lead with it. We publish the validated floor — with the raw number right beside it — because the gap between them is the story. Anyone quoting the gross number without the wash adjustment is quoting a ponzi.
Our adjustment is deliberately conservative and on-chain-verifiable:
localhost, test, or example hosts are excluded from validated.We measure at the settlement layer, forward-only, and disclose coverage: today we track roughly 77% of listed provider wallets by count (EVM + Solana). Solana and the rest are folded in the same way — neutrally, by measurement, no chain favored.
The agent economy is real, early, and small — and it will grow. When it does, the number that matters won't be the one in the press release. It'll be the one you can verify on-chain, with the wash stripped out and the coverage disclosed.
That's the number we publish, every day.
See the live figures on the Economy scoreboard, and how we measure at /methodology.
These numbers update continuously on the Economy scoreboard, with the methodology at /methodology.