Industry Analysis
Stripe, Visa, and Google All Shipped Agent Payment Protocols This Week. Here's What Merchants Need to Do.
Three protocols in seven days. If you're on Shopify, WooCommerce, or BigCommerce, you have a path. If you're on Magento, Miva, or NetSuite, you're not on any integration list.
Published March 2026 · 9 min read
TL;DR
- 1. Stripe MPP, Visa Intelligent Commerce, and Google UCP+AP2 all launched within seven days of each other
- 2. Shopify, WooCommerce, BigCommerce merchants are covered. Magento, Miva, NetSuite are not.
- 3. OpenAI killed ACP — the problem was never payments, it was real-time catalog and inventory data
- 4. Legacy platform merchants need 3 things: structured catalog, real-time inventory, protocol-compatible endpoints
- 5. 4-week readiness checklist at the bottom
The week of March 17, 2026: Stripe launched MPP (Machine Payments Protocol). Visa released Intelligent Commerce APIs. Google formalized UCP (Universal Commerce Protocol) and AP2 (Agent Payments Protocol), backed by 60+ organizations including Shopify, Etsy, Target, and Walmart.
AI agents can now discover products, negotiate terms, and complete transactions programmatically. The infrastructure exists. The question is whether your store can participate.
Stripe MPP: Machine-to-Machine Payments
Stripe's Machine Payments Protocol defines how AI agents initiate, authorize, and complete payments without a human clicking "Buy Now." Stripe positioned it as an open standard, not a proprietary API.
Early adopters:
Coach, Kate Spade, URBN, Etsy, Revolve, Ashley Furniture
Supported platforms:
Squarespace, Wix, WooCommerce, BigCommerce (plus Stripe-native merchants)
For merchants already on Stripe with a supported platform, the integration path exists. For merchants on Magento, Miva, or NetSuite, it does not. Stripe MPP supports the platforms it supports. If your store isn't on the list, you need a translation layer.
Visa Intelligent Commerce: Full Shopping Flow
Visa took a different approach — covering the full agent shopping flow, not just the payment: product discovery, offer application, payment method selection, and checkout. Visa CLI, built by Visa Crypto Labs, goes further: AI agents executing payments directly from a command line. No browser, no checkout page, no human in the loop.
100+ partners in sandbox · 30+ active integrations · 20+ live in production
APAC, Europe, LatAm rollouts planned mid-2026
Visa requires two things from merchants: your product catalog available via structured API, and pricing/promotions/inventory queryable in real time. An AI agent cannot work with "call for pricing" or "ships in 3–5 business days, maybe."
Google UCP + AP2: The Protocol Layer
Google's contribution is the most architecturally ambitious. UCP normalizes how AI agents communicate with merchants — a universal translator between any agent and any merchant, regardless of platform. AP2 handles the money: cryptographic proof of payment mandates so an AI agent can prove it has authorization to spend on behalf of a specific user.
Backed by Google, Shopify, Etsy, Target, Walmart, and 60+ other organizations. UCP is platform-agnostic in theory. In practice, the launch partners are Shopify and large marketplaces. A mid-market merchant on Miva with a 3-person IT team is not implementing UCP from scratch.
Protocol Comparison
| Dimension | Stripe MPP | Visa IC | Google UCP+AP2 |
|---|---|---|---|
| Scope | Agent-initiated payments | Full shopping flow + payment | Commerce protocol + payment auth |
| Supported platforms | Squarespace, Wix, WooCommerce, BigCommerce | API-based (platform-agnostic) | Shopify native; others via implementation |
| Magento/Miva/NetSuite | No | Not natively | Not natively |
| Production status | Live with early adopters | 20+ live integrations | Spec published, early implementations |
Why OpenAI Killed ACP
OpenAI's Agent Commerce Protocol was supposed to be the standard for how ChatGPT bought things. It didn't work out — but the reason matters.
The problem was not the payment flow. It was everything before it. ACP required merchants to maintain synchronized inventory and pricing data accessible to AI agents in real time. Most merchants couldn't deliver this. Product feeds are batch-updated. Pricing rules live in ERP systems without real-time APIs. Inventory counts in Magento or NetSuite lag behind warehouse reality by minutes to hours.
OpenAI shifted to an "Apps" model, pushing the inventory/pricing sync problem back to the merchant. The hard part of agentic commerce was never the payment. Stripe, Visa, and Google solved payments years ago. The hard part is getting a merchant's catalog, inventory, and pricing into a format that an AI agent can trust in real time.
What Legacy Platform Merchants Actually Need
If you run on Magento, Miva, NetSuite, or any platform not named in the announcements above: no one is building protocol support for your platform. You need a translation layer. That layer requires three things.
1. A structured, machine-readable product catalog
AI agents don't parse HTML product pages. They need structured data: product name, SKU, description, attributes, pricing, and availability in JSON-LD or via an API endpoint. For automotive merchants, this means fitment data too — year/make/model structured, not a PDF chart. Most Magento and Miva stores have this data in their database. It's not exposed in a way agents can consume. The gap is data accessibility, not data quality.
2. Real-time inventory and pricing
This is where OpenAI's ACP failed. Batch-updated feeds are not sufficient. Real-time means: if a product goes out of stock at 2:00 PM, the agent knows by 2:05 PM. For NetSuite merchants this typically requires a middleware layer polling inventory changes. For Magento, extending the REST API with inventory webhooks. For Miva, building an integration the platform doesn't natively support.
3. Protocol-compatible endpoints
At minimum: a product discovery endpoint returning structured catalog data, an availability endpoint returning real-time stock status, a pricing endpoint returning current pricing with promotions, and a cart/checkout endpoint accepting programmatic orders. These don't exist on Magento, Miva, or NetSuite out of the box.
4-Week Merchant Readiness Checklist
Week 1: Audit your catalog data
- □Export full product catalog — how many SKUs have complete structured attributes?
- □Identify SKUs with missing images, incomplete specs, or 'call for pricing' flags
- □For fitment-dependent products: is compatibility data structured or unstructured (PDF, footnotes)?
- □Run a free AI visibility audit — how do ChatGPT, Perplexity, Gemini see your products now?
Week 2: Map inventory and pricing infrastructure
- □How often does inventory data update? Real-time, hourly, daily, or manual?
- □Where does source-of-truth pricing live? ERP? Platform database?
- □Do you have API access to inventory and pricing systems, or are they behind admin UIs?
- □Document your current product feed process: who generates it, format, frequency, destination.
Week 3: Evaluate your API surface
- □Does your platform expose a REST or GraphQL API?
- □Can an external system query product availability by SKU in real time?
- □Can an external system create an order programmatically?
- □What authentication does your API use? Suitable for machine-to-machine access?
Week 4: Choose your integration path
- □Build internally: 3–6 months of engineering time to build and maintain protocol endpoints
- □Platform migration: move to Shopify/BigCommerce for native support — 6–12 months, significant risk
- □Middleware integration: connect existing platform to agent protocols — 1–2 weeks, $299/mo + 2% on AI-attributed revenue
Start with a free AI visibility audit
See how ChatGPT, Claude, and Gemini currently see your products. Takes 48 hours. No commitment.
Get your free audit →