ACP, AP2, UCP and x402: the agentic-commerce protocol map for store owners

ACP, AP2, UCP and x402 are not four competing standards you pick one of — they sit at three different layers of an agent-driven purchase and are built to stack, so the real question for a store owner is never "which one" but "am I ready for the layer underneath all of them." UCP and ACP handle how an agent discovers your products and completes a checkout, AP2 proves the shopper actually authorized the payment, and x402 moves the money; a single agentic sale can touch three of the four at once.

We spend our days on the one layer none of these protocols provides — the product data every one of them assumes is already clean — which is why we built Contexta, our WooCommerce plugin. ACP, UCP and the rest all take for granted that your catalog exposes a real price, a resolvable identifier and honest availability, and not one of them fills those fields in for you. So this map sorts the four by what they actually do, says which you can touch today, and marks where the hype and the reality still don't meet.

Are ACP, AP2, UCP and x402 competing standards you have to choose between?

No — they operate at three distinct layers and are designed to interoperate, so treating them as rivals is the single most common mistake in agentic-commerce coverage. Think of a normal card purchase online: HTTP carries the page, the checkout form collects intent, and Visa's rails move the money — nobody asks "should I use HTTP or Visa." The agentic stack splits the same way, and these four protocols slot into different rungs of it.

Discovery, cart & checkout

UCP (Google · Shopify) and ACP (OpenAI · Stripe) — how an agent finds your products and completes a purchase against your store

Payment authorization & trust

AP2 (Google, now FIDO Alliance) — cryptographically signed mandates proving the shopper approved this exact cart and price

Settlement — moving the money

x402 (Coinbase · Cloudflare) for stablecoins, alongside ordinary card rails — the actual transfer of funds

The clearest proof they're complementary came in June 2026, when Adyen shipped a single "agentic" integration that speaks UCP, ACP and AP2 at once — the payment industry isn't betting on one winner, it's building adapters that translate across all of them. For a store owner, that's the whole strategic lesson in one product launch: the layer you'd want to standardize on is the one below all four, not any single protocol above it.

What does each of the four protocols actually do?

Each solves a different piece of the problem, from three different sponsors, on three different timelines. ACP standardizes the in-chat checkout handoff; UCP standardizes the whole discovery-to-order journey; AP2 standardizes proof-of-authorization; x402 standardizes machine-to-machine settlement. Here is the map at a glance, with who backs each and where it stood as of mid-2026. One protocol that gets lumped into this conversation doesn't belong on this map at all: MCP, the standard behind WordPress's new adapter, is operator-facing — it lets your own agent run your store, not a shopper's agent buy from it.

ACP — OpenAI + Stripe

Checkout-execution layer. Launched Sept 2025; powered ChatGPT Instant Checkout, which OpenAI pulled in March 2026. The open protocol survives.

UCP — Google + Shopify

Discovery-to-order layer. Launched Jan 11, 2026; live on Gemini and AI Mode; Shopify opened self-serve agent registration in June 2026.

AP2 — Google → FIDO Alliance

Authorization layer. Sept 2025; three signed mandates (Intent, Cart, Payment); 60+ partners; governance handed to FIDO in April 2026.

x402 — Coinbase + Cloudflare

Settlement layer. Revives the HTTP 402 status code; 165M+ agent payments by April 2026; now a Linux Foundation project.

What unites all four is the assumption they never state out loud: that there is a structured, trustworthy catalog on the other end. UCP's agents read a manifest you publish at /.well-known/ucp; ACP's flow needs a feed; every one of them expects your products to carry the GTINs and barcodes that let an agent match your item to the same product elsewhere. A protocol endpoint over a catalog full of blank fields is a door into an empty room.

Did any of these protocols actually work in 2026?

One layer showed real volume and one showed real disappointment, and the split is instructive. x402 has genuine traction: by late April 2026 Coinbase reported roughly 69,000 active agents, 165 million transactions and about $50 million in cumulative volume, and the Linux Foundation formalized it as an independent project on April 2, 2026. Machine-to-machine payments for APIs and data — sub-dollar settlements where cards are too slow and expensive — turned out to be a real market.

The in-chat consumer checkout dream underdelivered. OpenAI shut down ChatGPT Instant Checkout in March 2026, roughly five months after launch, after fewer than 15 of Shopify's millions of merchants ever went live and sales stayed near zero — shoppers happily researched in ChatGPT but went back to a familiar store to buy. OpenAI pivoted to product discovery plus in-app merchant experiences, and the honest lesson for anyone weighing whether to chase a protocol is that a standard existing is not the same as buyers using it. That gap is exactly what to weigh before wiring your store into ChatGPT's shopping surfaces, where the discovery path is real but the direct-checkout path just proved fragile.

Which protocol should a WooCommerce store owner implement?

For almost every WooCommerce store, none of them directly — and that's not defeatism, it's how the layers actually reach you. Three of the four arrive through someone else: your payment processor adopts x402 and AP2, your checkout provider handles ACP's shared payment token, and the platforms that co-designed UCP flip it on for their own merchants. WooCommerce appears on none of the launch-partner lists, so a self-hosted store's only direct paths today are early third-party plugins sitting at 0.x versions, plus a Google Merchant Center dependency that matters more than the protocol code — which is why, across all four AI shopping surfaces, Google's is the one product-feed gate an independent store can still walk through unaided.

That's why "should I install a UCP adapter this quarter" is usually the wrong question, covered in full in what a WooCommerce store actually needs for UCP. The protocol layer will commoditize — within a year, support will likely be a checkbox from a matured plugin or WooCommerce itself. Spending scarce hours hand-wiring a 0.x adapter now, before agentic checkout drives measurable revenue anywhere, buys a small store very little.

If you can't pick a protocol, what should you actually do?

Fix the one layer you own and every protocol silently depends on: complete, consistent, machine-readable product data. This is the work that pays off no matter which standard wins, because ACP, UCP, AP2 and x402 all resolve to the same precondition — an agent has to be able to identify your product, trust its price, and confirm it's in stock before any checkout protocol has anything to execute. Get that right and you're portable across all four; get it wrong and none of them can transact you.

Concretely, that means a resolvable GTIN on every product that has one, a price that agrees across your page, your schema and your feed, real availability, and content an agent can read without running JavaScript — the same groundwork an AI shopping agent needs from your product record whether it arrives via Google, ChatGPT or a wallet-based agent paying in stablecoins. Doing that by hand across a few hundred SKUs is the task nobody finishes, which is why Contexta's Commerce Readiness audit checks each product for the fields agents actually need — image, price, GTIN, brand, weight — and hands back a per-product fix link. Protocols will keep launching, merging and dying; a catalog an agent can read and trust outlives all of them.