What is the difference between ACP, AP2, and x402?

They operate at different layers of an agent purchase. The Agentic Commerce Protocol (ACP), from OpenAI and Stripe, orchestrates a checkout inside a chat surface — discovery, cart, and buy in the conversation. AP2, from Google, is a payment-authorisation layer: it proves, cryptographically, that a human authorised what the agent is paying for. x402, stewarded by the Linux Foundation, is an HTTP-native payment-request standard — a server quotes a price and a client pays inside a normal web request. ACP is about completing an order, AP2 about authorising a payment, and x402 about carrying the payment request. They are complementary, not rivals.

Do I have to choose one agent-payment standard?

No. Because these standards sit at different layers, a single agent purchase can use more than one — orchestrated through a checkout standard, authorised through AP2, and settled over a card rail or through an x402 request. As of mid-2026 none has converged into a single winner, and the practical stance for a brand is to make your checkout able to accept however agents want to pay as these mature, rather than betting on one standard. The layer that actually decides revenue — whether an agent picks your product — sits above all three.

Which standard actually moves the money?

None of them, directly. ACP orchestrates the checkout, AP2 authorises the payment, and x402 carries the payment request — but the funds still settle on an underlying rail, either a card network or a blockchain (for stablecoins). This is a common point of confusion: the agent-payment standards define how an order is agreed, authorised, and requested, not the settlement itself. Cards and stablecoins are the rails; ACP, AP2, and x402 are layers built on top of them.

ACP vs AP2 vs x402: how the agent-payment standards compare

Three standards are being built for the moment an AI agent buys something, and they mostly sit at different layers rather than competing. The Agentic Commerce Protocol (ACP) orchestrates a checkout inside chat; Google's AP2 proves a human authorised the payment; x402 carries the payment request over HTTP. This is a neutral, criteria-based comparison of what each does, who backs it, and how they fit together — because for most brands the answer is not "pick one" but "understand which layer each owns."

All three are early and moving, so figures and adoption below are dated to mid-2026 and directional. The durable point is the layering, which is stable even as the details shift.

What are ACP, AP2, and x402, in one line each?

ACP (Agentic Commerce Protocol) — an OpenAI-and-Stripe standard that lets an agent complete a delegated checkout inside a chat surface like ChatGPT.
AP2 (Agent Payments Protocol) — a Google standard that attaches cryptographic proof of human authorisation to an agent-initiated payment.
x402 — a Linux Foundation standard that revives the HTTP 402 Payment Required status so a client can pay for a resource inside an ordinary web request.

Each solves a different piece of the same problem: how does a machine buy from you, provably and without a person clicking through a hosted checkout?

How do the three standards compare, criterion by criterion?

The criteria that actually distinguish them, compared even-handedly:

Criterion	ACP	AP2	x402
From	OpenAI & Stripe	Google	Coinbase → Linux Foundation
Layer	Checkout orchestration	Payment authorisation	Payment request (transport)
Core job	Complete an order inside chat	Prove a human authorised the payment	Ask for, and confirm, payment over HTTP
Key artefact	Delegated in-chat checkout	Signed Mandates (Intent, Cart, Payment)	The `402` challenge + proof-of-payment retry
Settlement rail	Via Stripe (cards)	Rail-agnostic (cards, bank, stablecoins via x402)	Rail-agnostic; mostly stablecoins today
Announced / stewarded	2025, OpenAI + Stripe	September 2025, 60+ partners	x402 Foundation, Linux Foundation (14 Jul 2026)
Best-fit scenario	Consumer buys inside an assistant	Any agent payment needing provable authorisation	Machine-to-machine, sub-cent, high-frequency payments

Sources: OpenAI/Stripe ACP materials; Google Cloud AP2 announcement (September 2025); x402 Foundation / Linux Foundation launch materials (July 2026). The pattern the table shows: the three answer different questions — complete, authorise, request — which is why they can stack rather than clash.

Are they competitors or complements?

Mostly complements, because they occupy different layers of one stack. Placed in order, an agent purchase can move through all of them:

Layer	Standard	What it decides
Discovery / being chosen	Answer-engine optimisation	Whether an agent surfaces and picks your product
Checkout orchestration	ACP	How the order is assembled and completed in chat
Payment authorisation	AP2	Whether the payment is provably human-authorised
Payment request	x402	How the payment is asked for and confirmed
Settlement rail	Cards or stablecoins	How the money actually moves

The agent-payment standards are not three roads to the same place. They are three layers of one road — orchestration, authorisation, and request — and the real contest sits above all of them, at whether an agent chooses you at all.

There is genuine overlap at the edges — AP2 explicitly uses the x402 extension for the stablecoin path, and both AP2 and ACP can front a card rail — but that overlap is interoperability, not rivalry. As of mid-2026 no standard has absorbed the others, and the field is still forming.

So which should a brand adopt?

For most brands, none of them is the thing to optimise for — and that is the point. These standards are settlement and authorisation plumbing; they move and validate money after an agent has decided to buy. Which product an agent surfaces and picks still turns on catalog richness, entity strength, reviews, and corroboration — the answer-engine optimisation that decides whether you are in the answer at all.

The practical stance: make your checkout ready to accept however agents want to pay as these standards mature, keep an eye on which gain real adoption, and spend your actual effort a layer up — on being the product agents choose. Whether AI agents and answer engines surface and recommend your products is exactly what Buffy Intel measures, snapshot over snapshot. Questions: [email protected].