Most Listed Paid Agent Services Have Not Yet Found a Second Customer

Of the 48,944 endpoints currently listed as paid services in the agent index, 12,434 have recorded more than one distinct paying buyer in the last 30 days. The rest have not. For a listed endpoint without a second buyer, the only transaction in its record is an operator running their own service.

The multi-payer count is the difference between a market and a demo. The data comes from Coinbase's call telemetry, not our own probe; it covers Coinbase's network and no other. A call count with one payer is evidence of an operator testing their own service; it says nothing about whether anyone else finds value in it. The index tracks both because the separation matters for any builder deciding whether to add a dependency: if you are integrating a paid endpoint, the question is not how many calls it has logged, but how many buyers have independently decided it was worth paying for.

On-chain settlement sharpens the picture further. In the trailing week, 321,838 USDC transfers reached agent payment addresses across 1,362 distinct services. That is real money moving to a much smaller set of services than the 48,944 listed. Concentration within that group is steep: www.ax1.vc alone accounts for 29.8% of all measured transfers.

There is no single reason a service would have calls but no second buyer. Some are new and have not been discovered. Some serve a use case narrow enough that the operator is the market. Some are simply unfit for production. The multi-payer threshold does not explain which situation applies; it tells you which question to ask before you find out the hard way.

The builder action is narrow and immediate: before writing any integration against a paid endpoint, query whether it has more than one buyer in the index. If it does not, treat it as a single-customer service that you are helping to become the second. That is a different risk posture than integrating into a market.