Which companies are building for agentic commerce?

Stripe, Shopify, J.P. Morgan, Google, commercetools, and Salesforce all offer agent-ready commerce infrastructure or protocols.

WebMCPAgentic CommerceE-Commerce

How WebMCP Powers Agentic Commerce: AI Agents Will Shop for Your Customers by 2030

Mukul Dutt

·March 19, 2026·12 min read

How WebMCP powers agentic commerce: AI agents will shop for your customers by 2030

Agentic commerce is the shift from humans clicking through online stores to AI agents handling the entire purchase flow. McKinsey projects it will generate up to $5 trillion in global revenue by 2030. Morgan Stanley estimates 126 million Americans will use AI shopping agents by then, covering 10 to 20% of all U.S. online retail.

This is not speculation. Amazon's Rufus already serves 300 million users. AI-sourced retail traffic surged 1,200% year-over-year while traditional search traffic declined 10%. And 88% of enterprises plan to modernize their commerce infrastructure within the next 12 months to prepare for agent-driven buying.

WebMCP sits at the center of this shift. It is a browser-level web standard that lets you register structured tools on your website, so AI agents can browse your catalog, compare products, and complete purchases through direct function calls instead of scraping your UI. If your store is not agent-ready by the time this wave hits, you will lose sales to competitors whose stores are.

The numbers behind agentic commerce

Market size forecasts

McKinsey calls agentic commerce a "seismic shift" with the same breadth of impact as the original web and mobile commerce revolutions. Their $1 trillion U.S. B2C projection and $3 to $5 trillion global figure by 2030 represent the high end, but even conservative estimates are massive.

Morgan Stanley projects $190 to $385 billion in U.S. e-commerce spending through agentic shoppers by 2030. Mordor Intelligence values the agentic AI in retail market at $60.43 billion in 2026, growing at 29.29% annually to $218.37 billion by 2031.

The B2B side is even bigger. Gartner predicts that by 2028, 90% of B2B purchases will be AI-agent intermediated, pushing over $15 trillion through automated exchanges. That is not a typo. Fifteen trillion dollars in B2B transactions handled by agents.

Consumer adoption is already here

You might think this is still years away. The data says otherwise.

Nearly 23% of Americans made at least one AI-assisted purchase in the past month, according to Morgan Stanley's AlphaWise survey. 58% of consumers have replaced traditional search with generative AI tools for product recommendations. And 76% say they want AI-powered shopping assistants.

The conversion numbers are what matter most. Shoppers who use AI chat convert at 12.3%, compared to 3.1% without. That is a 4x improvement. Customers interacting with AI complete purchases 47% faster. And returning AI-chat shoppers spend 25% more than those who skip the agent.

During Amazon Prime Day 2025, traffic from AI shopping tools jumped 3,300% year-over-year. Salesforce reported that online traffic from AI assistants grew 119% in the first half of 2025. These are not niche experiments anymore.

How agentic commerce actually works

The agent purchase flow

A customer tells their AI agent something like "order me running shoes under $120, same brand as last time, size 10." The agent already knows the customer's preferences from past purchases.

The agent searches across multiple stores. It checks inventory, compares prices, reads reviews, and applies any available coupons. It selects the best option based on the customer's criteria. Then it completes the purchase using a saved payment method with the customer's pre-set spending limits.

The customer gets a notification: "I found Nike Pegasus 41 at $109 from Store X. Arriving Thursday. I went ahead and ordered them since they match your preferences and are under your $120 limit."

That entire flow happens without the customer visiting a single product page.

Why traditional stores break

Here is the problem. Most online stores were built for humans with mice and keyboards. An AI agent trying to shop on a traditional e-commerce site has to render the page, parse the DOM, guess which buttons to click, fill in form fields by their CSS selectors, and handle pop-ups and CAPTCHAs.

That approach fails constantly. Websites redesign their layouts. JavaScript frameworks load content dynamically. A selector that worked yesterday breaks today. The agent wastes tokens rendering screenshots that cost 2,000 or more tokens each, when a structured tool call would use 20 to 100.

This is exactly the problem WebMCP solves.

How WebMCP makes your store agent-ready

WebMCP is a browser-level web standard developed by Google and Microsoft through the W3C. It shipped in Chrome 146 Canary in early 2026. Instead of letting agents scrape your UI, you give them structured tools to interact with your store directly.

Exposing your catalog as tools

With WebMCP, you register tools that agents can discover and call. A searchProducts(query, category, priceRange, sortBy) tool lets any agent search your catalog with structured parameters. A getProductDetails(productId) tool returns clean JSON with price, availability, sizes, colors, reviews, and images.

The agent does not need to render your product listing page. It calls your tool, gets structured data back, and uses that data to compare options across stores. Your product data arrives clean, complete, and in a format the agent can reason about.

Making checkout callable

The checkout side works the same way. You register a tool like addToCart(productId, quantity, options) and another like checkout(shippingAddress, paymentToken). The agent calls these functions directly instead of navigating through your cart page and multi-step checkout forms.

WebMCP's declarative API makes this simple for standard stores. You add three HTML attributes to your existing checkout form and it becomes an agent-callable tool.

<form toolname="checkout"
      tooldescription="Complete purchase with shipping address and payment."
      toolautosubmit="true"
      action="/api/checkout" method="POST">
  <input name="product_id" type="text" required>
  <input name="quantity" type="number" value="1">
  <input name="shipping_address" type="text" required>
  <input name="payment_token" type="text" required>
</form>

VentureBeat reported an 89% token reduction when agents use WebMCP tools versus screenshot-based approaches. Task accuracy hits 98% on structured tool calls. That efficiency gap is why WebMCP matters for commerce at scale.

The protocol stack powering agent commerce

WebMCP does not work alone. It is part of a larger protocol ecosystem that is coming together fast.

Four protocols, four jobs

MCP, the Model Context Protocol, is the foundation layer. It exposes data and tools from any system to AI models. WebMCP is its browser-facing implementation.

A2A, the Agent-to-Agent protocol from Google, handles communication between agents. When your shopping agent needs to negotiate with a store's sales agent, A2A manages that exchange.

ACP, the Agentic Commerce Protocol maintained by OpenAI and Stripe, standardizes product discovery, checkout, and payment capture specifically for commerce transactions.

UCP, Google's Universal Commerce Protocol co-developed with Shopify, powers checkout inside Google AI Mode and the Gemini app. Visa, Mastercard, American Express, Stripe, Target, Best Buy, and Macy's all endorsed it at launch.

The Linux Foundation established the Agentic AI Foundation with Anthropic, Block, Google, Microsoft, and OpenAI to drive interoperability across all of these protocols.

Where WebMCP fits

WebMCP is the layer between the agent and your website. The other protocols handle agent-to-agent communication, payment processing, and cross-platform coordination. But when an agent needs to interact with a specific store's catalog and checkout, it uses WebMCP tools.

Think of it this way. UCP lets a customer buy something inside Gemini. But the store still needs to expose its products and checkout flow in a structured way. WebMCP is how you do that on the open web, outside of any single platform's ecosystem.

Who is already building for this

The biggest names in commerce are moving fast.

Stripe and commercetools

Stripe launched its Agentic Commerce Suite in December 2025. It powers Microsoft Copilot Checkout, where users buy from Etsy, Urban Outfitters, and Anthropologie without leaving the chat. Onboarding brands include Coach, Kate Spade, Revolve, and Ashley Furniture.

commercetools launched AgenticLift, a standalone agentic offering that works without replatforming. Their AI Hub connects product and pricing data to ChatGPT, Copilot, and Gemini ecosystems. Frasers Group and Liverpool department stores already use it.

J.P. Morgan and Shopify

J.P. Morgan partnered with Mirakl in March 2026 to launch Mirakl Nexus for agentic commerce. J.P. Morgan handles payment tokenization and fraud protection. Mirakl provides catalog optimization and AI discovery. It is currently in closed beta with select retailers.

Shopify's numbers speak for themselves. AI-powered shopping orders jumped 15x since January 2025. AI-driven traffic to merchant stores grew 8x year-over-year. Stores optimized for agentic discovery convert 28% higher from AI traffic compared to traditional search.

Salesforce

Salesforce launched Agentforce Commerce to connect product catalogs and checkout flows to consumer AI channels like ChatGPT. 97% of commerce decision-makers surveyed by Salesforce agree that AI will reshape commerce.

What you should do now

You do not need to wait for 2030. The infrastructure is here and adoption is accelerating. Here is how to start.

Make your catalog discoverable

Add WebMCP tools that let agents search and filter your products. Start with your highest-traffic categories. A simple search tool with query, category, and price range parameters covers most use cases.

Expose checkout as a tool

Add the three WebMCP attributes to your checkout form if you use the declarative API. Or register an imperative checkout tool for complex flows with inventory checks, promotions, and shipping calculations.

Adopt scoped payment tokens

Integrate with Stripe's Shared Payment Tokens or a similar system. These tokens are scoped to specific sellers with time and amount bounds, so an agent can pay on behalf of a customer without accessing raw credit card data.

Optimize for agent discovery

Gartner warns that traditional SEO and PPC will give way to "agent engine optimization." Your products need to be machine-readable. That means structured data with clean tool descriptions, plus WebMCP tool schemas that clearly describe what each tool does and what parameters it accepts.

The stores that prepare now will capture the early wave of agent-driven commerce. The ones that wait will be invisible to the 126 million AI shopping agents Morgan Stanley projects by 2030.

What is agentic commerce?

Agentic commerce is a model where AI agents handle the full purchasing process on behalf of consumers. The agent searches, compares, selects, and buys products based on the customer's preferences and spending limits. McKinsey projects it will generate up to $5 trillion globally by 2030.

How does WebMCP enable agentic commerce?

WebMCP lets online stores register structured tools that AI agents can call directly. Instead of scraping product pages and checkout forms, agents call functions like searchProducts() and checkout() to interact with your store. This delivers 89% token reduction and 98% task accuracy compared to screen-scraping approaches.

Which companies are already building for agentic commerce?

Stripe powers agent checkout for Microsoft Copilot with brands like Etsy and Urban Outfitters. Shopify shipped MCP tools for developers. J.P. Morgan partnered with Mirakl for enterprise agentic commerce. Google launched UCP with backing from Visa, Mastercard, and 20 other companies. commercetools and Salesforce both offer agent-ready commerce platforms.

How big will agentic commerce be by 2030?

McKinsey projects $1 trillion in U.S. B2C revenue and $3-5 trillion globally. Morgan Stanley estimates 126 million AI shopping agent users in the U.S. alone. Gartner predicts 90% of B2B purchases will be agent-intermediated by 2028, representing over $15 trillion in transactions.

Do I need to rebuild my store for agentic commerce?

No. WebMCP's declarative API lets you add three HTML attributes to existing forms. Shopify's MCP server handles agent integration for Shopify stores. Stripe's payment tokens work with existing payment setups. You add agent support on top of your current infrastructure.