ProjectAtlas
Back to the presentation

Financial Applications

How Will AI Agents Buy and Sell Intelligence?

x402 makes it concrete: a request comes back priced instead of refused, the agent pays from an allowance a person approved, and the data is released in the same exchange. Settlement can run in stablecoin, in a native token where the rail supports it, or in a hybrid the treasury manages.

Main finding

Machine-to-machine payment is technically feasible today: a request can return a price, an agent can pay from a pre-approved allowance, and the data can be released in the same exchange. The open questions are demand and settlement design, not transport. Stablecoin, native-token and hybrid treasury settlement are three different designs with different legal exposure; purchase and burn are a fourth, optional layer.

What supports it

  • The x402 protocol defines how an HTTP request is answered with a payment requirement and retried once payment is attached.
  • The x402 flow is rail-agnostic: the payment requirement names the asset and network, so stablecoin and native-token settlement use the same request path.
  • Agents already take multi-step actions and call external tools, which is the client side of such a purchase (Yao et al. 2022).
  • Machine-readable output is a product decision: a report built for an agent looks different from one built for a person.

What this does not claim

  • Demand is unvalidated. No customer, price point or volume is evidenced here.
  • Accepting machine payments is a regulated activity in many jurisdictions and needs legal review before a pilot.
  • No price appreciation, yield or return is implied by any settlement design described here; a burn rule changes supply, not demand.

Planning assumptions used

  • The first product is a narrow, verifiable feed (liquidity or risk), not a broad intelligence platform.
  • Payments settle in stablecoin on a supported rail for the pilot.
  • Any token layer is built after the intelligence service has paying usage, not before it.

Open before signature

  • Which single dataset would a paying machine actually buy first?
  • What price per request is defensible against free alternatives?
  • Which entity holds the payment licence, and who bears refund risk?
  • Which settlement design survives if the token is removed entirely?

Sources

Figures marked illustrative are planning assumptions. This note separates published evidence from project assumptions and does not present a supplier quotation, a valuation or a forecast of returns.