Forty-eight hours after Klarna plugged its merchant network into ChatGPT, MoonPay launched as the first crypto onramp inside the same surface. Users can now buy crypto without leaving the conversation. Two integrations, one week, two financial primitives. The pattern is not subtle.
We think OpenAI is building a financial app. The press releases call it an App Store. The economics call it something else.
When the same surface ships BNPL, crypto onramp, and (in a previous attempt) checkout within sixty days, the integrations stop being a product story and become a positioning story.
What actually shipped
MoonPay announced on May 22 that its onramp is live inside the ChatGPT App Store, as reported by PYMNTS. A user searches "MoonPay" inside ChatGPT and completes a crypto purchase in-chat, with funds settling to the wallet of their choice. MoonPay's framing: "We are officially the first and only crypto onramp integrated in ChatGPT."
The integration matters more than its individual capability. MoonPay processes around $4 billion in annual onramp volume, which is meaningful but not market-shifting on its own. The shift is what it joins inside the same chat surface.
Two days earlier, Klarna embedded its merchant network inside ChatGPT so that users could buy products from millions of merchants without leaving the conversation. Two months before that, OpenAI tried and walked back its own checkout product for merchants. Three financial primitives, discovery, buy-now-pay-later, and crypto onramp, converging on a single chat surface in a matter of weeks.
The Klarna and MoonPay pattern
The two integrations look different at first read. Klarna is consumer credit. MoonPay is crypto rail. They sit at opposite ends of the payments stack.
Read them together and the pattern is obvious. Both deliver a complete financial transaction inside ChatGPT. Both replace a user journey that previously required leaving the chat to visit a separate destination. Both put OpenAI as the surface owner without OpenAI taking the regulated payments role itself.
As Simon Taylor has been writing, the new commerce stack is vertical rather than horizontal, with surfaces stacking on top of rails rather than competing alongside them. ChatGPT is the surface. Klarna and MoonPay are the rails. OpenAI does not need a banking license to be the most important name on every receipt.
This is the operating logic. The integrations sit on the MM Trust Layer Model, with OpenAI controlling the discovery layer and routing the consumer's standing intent into whichever partner owns the authorization, payment, and settlement layers. Klarna handles BNPL settlement. MoonPay handles crypto rail. OpenAI handles the relationship.
What changes for incumbent payments players
Three groups have to reread their strategy this morning.
Card networks. The MoonPay integration routes consumer purchases into crypto. The Klarna integration routes consumer purchases into BNPL. Neither uses traditional Visa or Mastercard rails directly. The card networks are not absent from agentic commerce, but every additional rail that lands inside ChatGPT is a transaction that does not necessarily touch their interchange. Visa Trusted Agent Protocol and Mastercard's agent moves are the network response. They have not yet shown up inside ChatGPT itself.
Exchanges and brokerages. Coinbase, Kraken, Robinhood, and every consumer crypto destination just lost a piece of the discovery layer. A user who would have downloaded Coinbase to buy USDC now buys USDC in ChatGPT. Coinbase's response will be either a deeper partnership (likely) or an OpenAI bypass via Coinbase's own AI agent (also likely). One of those resolves into product positioning, the other into a partnership race.
Banks. The integration also gives ChatGPT a structural reason to know each user's standing payment preferences, wallet addresses, and crypto holdings. That is a data position banks have spent decades protecting. As we framed in the agentic commerce practitioner definition, the discovery layer is where the customer relationship lives now. Banks holding the deposits while OpenAI holds the relationship is a structural disadvantage banks will have to address with their own agent surfaces.
The destination economy thesis, restated
We wrote two days ago that Klarna picked the agent surface over its own storefront. We frame this pattern as the MM Destination Economy Thesis: when the shopper is an agent, every owned-destination strategy depreciates.
MoonPay is the strongest confirmation yet. MoonPay's primary growth channel for years has been embedding inside other apps and platforms. The integration into ChatGPT extends a strategy MoonPay already runs. What is new is the surface MoonPay chose, and the implicit positioning that ChatGPT now competes with Coinbase, Robinhood, and Cash App for crypto-buying eyeballs.
The destination economy is not collapsing all at once. It is being unbundled into surfaces. ChatGPT is one of the largest emerging surfaces. The next 12 months of integrations will determine whether it is the dominant one or one of three.
The regulatory question OpenAI is not answering
A point coverage will gloss over: at what scale does OpenAI itself become a regulated payments entity?
Today, OpenAI does not hold consumer funds. Klarna underwrites BNPL. MoonPay handles crypto rail. OpenAI takes a (likely) revenue share and routes the user. Under existing payment service provider rules in the US and EU, that is currently fine. ChatGPT is the surface, not the regulated party.
The line moves if OpenAI ever holds value, even briefly, between user authorization and merchant settlement. Or if it issues its own stablecoin. Or if the integrations expand into payments-out (not just onramp). Each of those would trigger payment service provider obligations in multiple jurisdictions.
Our forecast (see the MM Theses workspace): within 24 months a major regulator opens a consultation specifically on agent-platform payments. The MoonPay integration accelerates that timeline because it is the first integration where consumer value passes through an LLM surface in a regulated asset class.
What to watch in the next 30 days
Three things.
First, which financial partner integrates next. Likely candidates: a major US bank (Chase has the scale; Capital One has the agility), a stablecoin issuer (Circle is the obvious one, given recent partnership posture), or a stockbroking layer (Robinhood would be the marquee). Any of these in the next 30 days is a structural signal.
Second, whether OpenAI publishes a developer SDK specifically for financial app integration. The current App Store route is one-off partnerships. An SDK is the platform play. If we see an SDK, the next year is consolidation; if we do not, OpenAI is still curating its rails one partnership at a time.
Third, the Visa or Mastercard response. The networks have published Trusted Agent Protocol and Multi-Token Network. Neither has shipped inside ChatGPT. The first network protocol that ChatGPT consumes natively is a market-moving event.
Sources
If ChatGPT is the financial app, who is the regulator?
Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.