Seat license vs pay-per-call: what changes when your API buyer is an agent
A seat license prices for a human looking at a screen. An agent never looks at a screen. Here is what breaks, and what pay-per-call access looks like instead.
The short answer
A seat license is priced per person, on the assumption that one human reads the data on one display. An agent that reads many sources and passes the output downstream is, in the language of the old licenses, a non-display machine consumer, and the seat model has no row for it. Pay-per-call access prices the request instead of the person, which is the unit an agent actually wants: one call, not a monthly chair.
Where the seat model breaks
Every price in the data business has a person hidden inside it. Non-display and redistribution clauses exist to stop that person from turning a feed into something machine-readable that flows past the seat. That structure worked for decades because the consumer really was a viewer. It stops working the moment the consumer is a program that never renders a screen.
People inside the data industry are saying this out loud: when the consumer is no longer human, licensing has to move toward machine identity, metering, and entitlement as a control layer, rather than counting seats. That pressure lands hardest on the long tail, quant shops, emerging managers, fintechs, independent researchers, who need data from many providers and cannot economically negotiate a per-seat contract with each one.
What pay-per-call looks like in practice
Take the seat out and price the call. An agent holds one identity and one prepaid balance, pays per request at a price it sees before it spends, and gets a signed receipt back on every call. That receipt matters more than it sounds: once something is spending money on your behalf across many providers, what you actually need is one verifiable record of what it bought, not ten dashboards to reconcile.
Proxygate runs this model today: one prepaid USDC balance, pay-per-call pricing, and per-call signed receipts, across a real-world data catalog that is still being filled out provider by provider. Two months live, so this is a description of what works now, not a roadmap slide.
What is still ahead
Turning per-call receipts into something a regulated data owner would accept as a legitimate way to reach agents is a harder problem, and it is ahead of this, not behind it. Worth saying plainly rather than implying it is already solved.
Related concepts
More guides
Frequently asked questions
A seat license prices for one human viewing data on one screen. An agent reads many sources and passes results downstream without ever displaying anything, so it does not fit the unit the license was built around.
Not the largest funds, who already hold enterprise licenses. It lands on the long tail: quant shops, emerging managers, fintechs, and independent researchers who need cross-provider data but cannot economically negotiate a seat contract with every source.
One identity and one prepaid balance per agent, a visible price before each request, and a signed receipt after it. The call becomes the billing unit instead of the person.