# There Are Now 15 Competing Agentic Payment Standards > ACP, UCP, x402, KYAPay, and counting. Why we refuse to build the 15th standard, and route on all of them instead. - Published: July 7, 2026 (2026-07-07) - Reading time: 5 min - Canonical: https://onecheckout.ai/blog/15-competing-standards - HTML version: https://onecheckout.ai/blog/15-competing-standards --- You know the [xkcd](https://xkcd.com/927/). Situation: there are 14 competing standards. "Ridiculous! We need to develop one universal standard that covers everyone's use cases." Soon: there are 15 competing standards.
xkcd comic #927, Standards: the joke that proposing one universal standard to unify 14 competing standards simply yields 15 competing standards.
Comic by xkcd (“Standards”, #927), used under CC BY-NC 2.5.
Agentic payments is speedrunning the comic. In roughly eighteen months the industry has produced ACP from OpenAI and Stripe, UCP and AP2 from Google, x402 from Coinbase, KYAPay from Skyfire, Visa Intelligent Commerce and the Trusted Agent Protocol, Mastercard Agent Pay, PayPal Agent Ready, and Stripe Shared Payment Tokens. We count at least ten, and it's Tuesday. Every one of them was announced as the way agents will transact. Several of them are genuinely good. And if you're building an agent, or a platform, or anything that needs to buy something this quarter, the roster is not a menu. It's a problem. So here is our position, stated plainly enough to be held against us later: **we are not building the 15th standard.** We read the comic as a warning, not a business plan. ## Three layers, usually conflated Most confusion about these protocols comes from treating them as competitors when they mostly aren't. They live at three different layers of the transaction. **Merchant protocols** (ACP, UCP) standardize how a merchant exposes catalog, checkout and order state to an agent platform. They're clean and well designed, and they share a dependency: the merchant has to adopt them, integrate feeds, and keep pricing and inventory synchronized for a channel that barely has volume yet. Adoption is the whole game, and adoption has been slow. OpenAI scaled back native ChatGPT checkout this spring after a handful of merchants went live. That's not an indictment of the design. It's the physics of asking millions of merchants to do work. **Network tokens** (Mastercard Agent Pay, Visa Intelligent Commerce, PayPal Agent Ready, Stripe Shared Payment Tokens) work at a different layer entirely. They're payment credentials that carry verified agent identity on the rails merchants already accept, at the checkout forms merchants already have. No catalog work, no feed maintenance. The card networks are distributing them on the largest payment install base on earth. These aren't competing with ACP and UCP so much as solving a different sentence in the same paragraph. **Alternative rails and identity** (x402, KYAPay) come from outside the card system. x402 revives the HTTP 402 status code for machine-speed stablecoin payments, and it's found real traction where card economics simply fail: sub-dollar metered API calls and agent-to-agent transactions. KYAPay pairs a payment token with Know Your Agent identity so the internet's bot defenses can tell a legitimate agent from a scraper; Skyfire has published it openly and submitted it to the IETF. Both are doing useful work in territory the card rails can't serve. Notice what's missing from all three layers: the ability to buy from a merchant who has adopted nothing. Which is, today and for years to come, almost all of them. ## The zeroth standard Here's the thing the standards conversation keeps stepping over. A universal agentic commerce protocol already exists. It has near-total merchant adoption, decades of fraud tooling, a dispute process your grandmother has used, and a rewards program attached. It's the credit card and the checkout form. The most widely deployed agentic payment standard is the HTML card field. Nobody announced it, so nobody counts it. That's why OneCheckout is built universal-first. Our routing engine is a cascade of eight methods that scores every transaction on viability and cost. Native agentic tokens sit at the top, because when Agent Pay or Visa Intelligent Commerce can carry a purchase, that's the best version of the transaction: verified identity, the buyer's own card, network-grade authorization. Below them sit the universal methods that work on the web as it actually exists, down to a wrapped fallback for the stubborn last percent. The buyer's card stays the buyer's card the whole way down, points and dispute rights intact. A cascade is the opposite of a standard. A standard says: transact my way. A cascade says: we'll transact whichever way works, ranked by what's best for the buyer. ## What we do with each of them This is the part of our FAQ that said "watch this space," expanded. **ACP and UCP:** where a merchant speaks them, we'll happily listen. UCP's structured data is useful for offer resolution even when the purchase itself routes elsewhere, and our intention is to layer it in where it helps. We just won't make our coverage wait on merchant adoption curves we don't control. **The network tokens:** these are our preferred rails, full stop. Every one that ships gets evaluated for the cascade, and the ones that are live are already in it. When we say virtual cards are hacks, this is the alternative we mean. **x402:** the honest gap in card rails is the microtransaction, and stablecoin settlement fills it. Our plan is a card-to-stablecoin bridge, so a buyer's ordinary credit card can fund x402-priced services without the buyer needing to know what a wallet is. **KYAPay:** identity is complementary to everything above. If the bot-management vendors Skyfire has assembled start honoring agent passports, our agents will be glad to carry one. Their success makes our universal methods more reliable and costs us nothing. The pattern is the same in every case. We don't compete with standards. We consume them. Each new protocol that earns its place becomes another method in the cascade, and the integration you did with us doesn't change. That's the quiet economics of universal-first: every standards announcement makes OneCheckout better, and none of them is load-bearing. ## Soon: there are still 8 methods The comic's punchline is that the unifier becomes the fragmenter. The way out isn't a better 15th standard. It's refusing the premise that buyers and sellers should wait for consensus before agents can transact. So our pledge is the boring one. When standard number 16 ships next quarter, we'll read the spec, argue about it in Slack, and if it's good, it becomes method 9. Your agent won't notice, except that slightly more of the internet works. Your agent should be able to buy from the internet that exists, not the internet the working groups are drafting. That's the whole thesis. The rest is routing. **[Try the free MCP →](https://onecheckout.ai/connect)**