Live|Chrome 146 — Feb 9, 2026
Google Just Changed
How People Shop Online
WebMCP lets AI agents search your products, add to cart, and check out using structured tools instead of screen-scraping.
Stores with better tool definitions give agents more reasons to transact there.
Google & Microsoft standardHow it works
AI Agent → Your E-Commerce Store
Without WebMCP
// Agent scrapes product page
✗Finding search input... not found
✗Parsing size dropdown... wrong element
✗Clicking "Add to Cart"... button moved
⚠Added size "L" instead of "10"
Result: Wrong item. Customer returns it.
With WebMCP
// Agent calls structured tools
✓search_products({query: "running shoes", maxPrice: 100})
✓3 results with price, size, stock
✓add_to_cart({sku: "NK-1234", size: "10", color: "blue"})
✓Cart updated. Exact match.
Result: Right item. Happy customer. Agent prefers your store.
02 / Context
Shopping Is Becoming a Protocol, Not a Page Visit
Three eras of how customers find and buy products. The third is happening now.
2000s
Browse & Click
Customers visit your site. They browse. They click. They fill forms. You optimize for eyeballs and page views.
2010s
Swipe & Tap
Customers discover via Instagram, TikTok, Google Shopping. You optimize for thumb-friendly UI and ad targeting.
2026
Ask & Transact
Customers tell their AI agent what they want. The agent calls your store's WebMCP tools — structured, typed, reliable. No browsing. No form-filling. Just: intent → tool invocation → transaction.
OPTIMIZE FOR: TOOL QUALITY + AGENT PREFERENCE
What This Means for E-Commerce
WebMCP is a proposed web standard released by Google and co-authored with Microsoft. It's available today in Chrome 146. It lets websites expose their functionality as structured tools that AI agents can discover and call — with typed parameters, JSON schemas, and reliable responses.
Product search
A text box an agent has to guess how to use
A callable function with typed filters
Add to cart
A button an agent has to find and click
A structured API call with exact SKU
Checkout
A fragile multi-step form navigation
A confirmed, auditable transaction
This isn't a chatbot. This isn't a plugin. This is a browser-level protocol that makes your store's functionality directly accessible to any AI agent running in Chrome.
0
browser share
StatCounter, 2026
0
open standard
WebMCP specification
0
proprietary lock-in
WebMCP is an open web standard
03 / How It Works
See What Agent-Ready E-Commerce Actually Looks Like
Two tracks. Pick the one that speaks your language.
Step 1
Your Customer Speaks
"Hey, find me running shoes under $100 in size 10"
A person talking to their AI assistant with a natural language request.
Step 2
The Agent Discovers Your Tools
The agent sees your store has registered WebMCP tools — structured functions it can call directly.
search_products • add_to_cart • checkout
Step 3
The Agent Searches Your Catalog
It sends a structured request: product type, max price, size. Your store returns exact matches with prices, images, and ratings.
Structured request → structured response with product cards.
Step 4
The Agent Adds to Cart
"I'll take the blue Nike Air Max." The agent calls add_to_cart with the exact SKU, size, and color. No wrong sizes.
Precise cart addition with confirmation.
Step 5
Secure Checkout
The agent initiates checkout. Because this involves payment, WebMCP triggers a human confirmation step — the customer reviews and approves.
requestUserInteraction → customer confirms → order placed.
The entire interaction takes seconds. The customer never opened your website. But they just bought from you — accurately, securely, and with zero friction.
04 / Tool Catalog
The 8 Tools Every Store Needs
WebMCP turns your store into a structured API that AI agents call directly. Each tool maps to a real shopping action.
PRODUCT DISCOVERY
—The "Find It" Stage
3 toolssearch_products()“Find me running shoes under $100”
Agent calls your catalog with structured filters — keyword, price range, size, category. Returns matching products with price, availability, images, and ratings.
The #1 tool agents call. If your search returns clear, structured results, agents prefer your store. If it doesn't exist, agents skip you.
get_product_details()“Tell me more about this shoe”
Agent passes a product ID and gets full specs — description, materials, sizing guide, reviews summary, related products.
A thin response loses to a competitor whose tool returns materials, fit notes, and ratings from 2,300 reviews.
browse_category()“Show me what's on sale”
Agent navigates your collections — sale items, new arrivals, seasonal — with pagination and sort options.
Not every shopping trip starts with a specific product. Browsing tools let agents serve "just looking" customers.
TRANSACTION
—The "Buy It" Stage
3 toolsadd_to_cart()“Add the blue one in size 10”
Agent passes exact SKU, size, color, quantity. Cart updates. Structured enum parameters mean zero wrong-size errors.
Cart errors are the #1 source of agent-driven returns. Typed parameters eliminate guesswork.
apply_coupon()“Use my SAVE20 code”
Agent passes coupon code, gets back discount amount and updated cart total. If invalid, returns clear error.
Coupon application is a trust signal. Agents that reliably apply discounts build customer loyalty.
checkout()“I want to buy this”
Agent initiates checkout with shipping and payment selection. requestUserInteraction() triggers human confirmation.
The revenue moment. Uses requestUserInteraction() for payment — a WebMCP security requirement.
POST-PURCHASE
—The "Track It" Stage
2 toolstrack_order()“Where's my package?”
Agent passes order ID or email, gets shipping status, tracking number, estimated delivery date.
Every "where's my order" handled by an agent is one fewer support ticket.
initiate_return()“I need to return these shoes”
Agent passes order ID and reason, gets return label URL and refund timeline. Uses requestUserInteraction().
Easy returns build trust. Agent-friendly return flows mean customers just ask their AI assistant.
All 8 tools auto-generated from your existing store pages. The Co-Pilot scans product forms, cart flows, and checkout to produce production-ready WebMCP definitions.
Scan My Store05 / Platform Support
Plug Into Your Existing Platform
Shopify, WooCommerce, or custom build — there's a path that matches your tech stack.
Shopify
FeaturedAgent-ready in 5 minutes. No code required.
1
5 minInstall from App Store
One-click install. Approved by Shopify.
2
2 minAI Co-Pilot Scans
Finds 4 product collections, 1 search form, 1 cart page, 1 checkout. Generates 8 tools.
3
15 minReview & Customize
Review each tool. Customize descriptions. Toggle on/off. Zero technical knowledge required.
4
1 clickGo Live & Monitor
Deploy with one click. See agent traffic within hours. Track every interaction and conversion.
Install on Shopify
Designed to map onto Shopify's product catalog structure with minimal configuration.
WooCommerce
WordPress plugin. Full WooCommerce integration.
- Auto-detects WooCommerce product structure
- Works with popular themes and page builders
- Compatible with WooCommerce Blocks and classic checkout
Magento / Adobe Commerce
Full-featured extension.
- Supports Magento 2.4+ and Adobe Commerce
- Multi-store, multi-language tool generation
- Enterprise security compliance built in
BigCommerce
Native integration via BigCommerce API.
- Stencil and headless storefront support
- Channel-aware tool definitions
- Full catalog, cart, and checkout coverage
Custom / Headless
JavaScript SDK for any architecture.
- npm package: @webmcp/ecommerce
- React/Next.js hooks: useAgentTools()
- Full TypeScript support with e-commerce types
06 / Intelligence Layer
Every Feature Solves a Real Agent Problem
7 capabilities built on WebMCP protocol mechanics — each solving a specific agent-commerce failure mode.
Problem
Solution
Impact
Problem
Agent recommends an out-of-stock product. Customer is frustrated. Agent learns to avoid your store.
Real-Time Inventory Awareness
WebMCP Solution
Tool responses return live stock levels. Agent sees { inStock: true, quantity: 23 } and only recommends available items.
Business Impact
Agents can check availability before recommending. Over time, stores with accurate stock data should become preferred.
WebMCP Solution
Tool responses return live stock levels. Agent sees { inStock: true, quantity: 23 } and only recommends available items.
Business Impact
Agents can check availability before recommending. Over time, stores with accurate stock data should become preferred.
Problem
Customer says "blue in large." Agent guesses form fields, selects wrong dropdown. Wrong item arrives.
Structured Variant Handling
WebMCP Solution
add_to_cart uses inputSchema enums: { size: { enum: ["S","M","L","XL"] } }. Agent picks valid options only.
Business Impact
Agents can only select valid options, which should significantly reduce wrong-variant orders from agent-initiated purchases.
WebMCP Solution
add_to_cart uses inputSchema enums: { size: { enum: ["S","M","L","XL"] } }. Agent picks valid options only.
Business Impact
Agents can only select valid options, which should significantly reduce wrong-variant orders from agent-initiated purchases.
Problem
Agent quotes cached price. Customer arrives at checkout — price is different. Trust destroyed.
Dynamic Pricing Transparency
WebMCP Solution
Every tool response includes real-time pricing. Sales, member pricing, and bulk discounts all reflected live.
Business Impact
Price accuracy builds agent trust. Agents prefer stores with consistently correct pricing.
WebMCP Solution
Every tool response includes real-time pricing. Sales, member pricing, and bulk discounts all reflected live.
Business Impact
Price accuracy builds agent trust. Agents prefer stores with consistently correct pricing.
Problem
Agent can't answer "When will it arrive?" because shipping is calculated in JavaScript the agent can't execute.
Agent-Aware Shipping
WebMCP Solution
get_shipping_options tool accepts address + cart, returns structured options with prices and dates.
Business Impact
Shipping transparency is the #2 factor in purchase decisions. Agents that answer delivery questions close more sales.
WebMCP Solution
get_shipping_options tool accepts address + cart, returns structured options with prices and dates.
Business Impact
Shipping transparency is the #2 factor in purchase decisions. Agents that answer delivery questions close more sales.
Problem
Your recommendation engine only works in the UI. Agents can't see JavaScript-rendered widgets.
Intelligent Product Recommendations
WebMCP Solution
get_product_details returns a relatedProducts array. Agent naturally suggests: "You might also like..."
Business Impact
Agents can surface related products naturally, recovering cross-sell opportunities that are invisible in screen-scraping workflows.
WebMCP Solution
get_product_details returns a relatedProducts array. Agent naturally suggests: "You might also like..."
Business Impact
Agents can surface related products naturally, recovering cross-sell opportunities that are invisible in screen-scraping workflows.
Problem
International customer asks agent to shop in local currency. Agent sees USD, can't convert.
Multi-Currency & International
WebMCP Solution
Tools accept locale/currency parameter. Responses return prices in requested currency with proper formatting.
Business Impact
International agents can present prices in local currency, removing a friction point that often blocks cross-border agent transactions.
WebMCP Solution
Tools accept locale/currency parameter. Responses return prices in requested currency with proper formatting.
Business Impact
International agents can present prices in local currency, removing a friction point that often blocks cross-border agent transactions.
Problem
You can't tell how much traffic is human vs agent. Your funnel data is polluted.
Agent Traffic Analytics
WebMCP Solution
Analytics SDK detects SubmitEvent.agentInvoked — the only reliable signal for agent-initiated actions.
Business Impact
Clean data. Optimize for TWO channels (human + agent) instead of one muddy channel.
WebMCP Solution
Analytics SDK detects SubmitEvent.agentInvoked — the only reliable signal for agent-initiated actions.
Business Impact
Clean data. Optimize for TWO channels (human + agent) instead of one muddy channel.
07 / Evidence
Evidence & Projected Impact
WebMCP is new. We share our reasoning transparently: protocol mechanics, verifiable market context, and honest projections.
A Note on Honesty
WebMCP is a brand-new protocol. Nothing below is based on production data at scale. We share reasoning from protocol mechanics, how the protocol works, and verifiable market context. As the ecosystem matures, we'll replace projections with real-world measurements.
Protocol Reasoning
Protocol Mechanic
Market Context
Projected
Protocol Reasoning
Structured Data Should Outperform Scraping
When agents receive typed JSON with product data, pricing, and availability — instead of parsing arbitrary DOM elements — transaction reliability should improve significantly. The same principle that makes APIs more reliable than screen-scraping applies here.
Reasoning from protocol mechanics: typed responses vs unstructured HTML parsing
Projected from protocol design, not yet measured in production
Protocol Mechanic
Quick Setup for CMS Platforms
For stores on platforms like Shopify, WebMCP tool definitions can be generated from existing product catalog structures. The protocol is designed to map onto standard e-commerce patterns (search, cart, checkout) without requiring custom development.
Based on WebMCP specification tool registration API
Projected
Protocol Reasoning
Potential for a Major New Traffic Channel
Chrome holds 65%+ global browser market share. With WebMCP shipping in Chrome 146 and Microsoft backing the spec, agent-mediated e-commerce could grow into a meaningful sales channel — though adoption timelines remain uncertain.
StatCounter Global Stats + Chrome release roadmap
Forward-looking estimate; actual adoption will depend on consumer and agent behavior
Protocol Mechanic
Typed Parameters Reduce Selection Errors
When variant selection uses inputSchema enums (e.g., size: { enum: ["S","M","L","XL"] }), agents can only choose valid options. This eliminates the class of errors where an agent guesses the wrong dropdown value from unstructured HTML.
WebMCP specification: inputSchema with JSON Schema enums
Projected
Protocol Reasoning
Post-Purchase Tools Could Deflect Support Load
If AI agents can call track_order and initiate_return tools directly, they could handle common support queries without human intervention. The potential depends on what share of support volume comes from these predictable request types.
Reasoning from typical e-commerce support ticket distribution
Projected benefit; actual impact depends on implementation quality and agent adoption
Market Context
Open Standard Backed by Browser Vendors
WebMCP is co-authored by Google and Microsoft, shipping in the browser with the largest market share. This is not a proprietary API or third-party plugin — it is a proposed web standard with browser-level integration.
WebMCP specification, Chrome 146 release notes
08 / Scoring
Your Agent Readiness Score
Like Google PageSpeed for AI agents. One number that tells you how well agents can shop your store — and how to improve it.
Sample Store Score
Based on a typical Shopify store
Score Dimensions
▾ = benchmark targetAre your forms annotated as WebMCP tools with proper inputSchema, descriptions, and types?
- search_products tool with keyword, price, category filters
- add_to_cart with typed variant/size/color enums
- checkout with requestUserInteraction() security
- All tools have clear human-readable descriptions
Meets benchmark (80)
Do your tool responses return structured, complete data that agents can work with?
- Product data includes price, availability, images, ratings
- Shipping options return dates and costs, not just labels
- Error responses are clear and actionable
- Related products included in product detail responses
3 points below benchmark (75)
Is payment handled via requestUserInteraction() per WebMCP security spec?
- Checkout uses requestUserInteraction() for payment
- Returns/refunds require human confirmation
- Cart modifications don't auto-charge
- Session management prevents phantom orders
Meets benchmark (85)
Are prices, stock levels, and shipping estimates returned in real-time?
- Prices reflect current sales and promotions
- Stock levels updated per-request, not cached
- Shipping estimates calculated live with address
- Currency/locale handled per-request
5 points below benchmark (70)
Quick Score
60 sec • No signupStatic analysis of your HTML. Detects forms, inputs, product structured data. Estimates tool potential without installing anything.
- No login required
- Scans public pages
- Estimates tool count
- Identifies quick wins
Recommended
Deep Score
5 min • Free accountAI Co-Pilot crawls your site. Tests form interactions. Validates tool definitions end-to-end. Produces actionable implementation plan.
- Authenticated scan
- Tests all forms + flows
- Generates tool definitions
- Prioritized action plan
Run a Free Store Audit
No signup required for Quick Score. Deep Score requires free account.
09 / Competitive Intel
Competitive Intelligence
An illustrative look at how agent readiness could differentiate your store. Scores below are examples, not real competitor data.
agent-readiness-dashboard (illustrative)
Example Category View
Agent Readiness Scores (illustrative)
Your Store (example)
0
Store without WebMCP
Not agent-ready
Store with partial setup
0
Store without WebMCP
Not agent-ready
Tool Capability Comparison (illustrative)
| Store | Search | Cart | Checkout | Tracking | Tools |
|---|---|---|---|---|---|
| Your Store (example) | 8 | ||||
| Store without WebMCP | 0 | ||||
| Store with partial setup | 1 | ||||
| Store without WebMCP | 0 |
Illustrative: How Agent Preference Could Work
This is a conceptual model, not measured data.
Full WebMCP store
72%
Partial implementation
18%
No WebMCP
10%
Illustrative model. When multiple stores sell similar products, agents are expected to prefer stores with better tool definitions, response quality, and transaction completeness. Actual preference patterns will depend on agent implementation.
Why Stores Without WebMCP Lose Agent Traffic
1
No product search tool
Agents cannot discover products — they have no way to query your catalog
Add search_products with keyword, price, category parameters
2
Unstructured variant selection
Agents may select the wrong size or color when guessing from unstructured form fields
Use inputSchema enums for size, color, and variant options
3
No shipping information tool
Agents cannot answer delivery timeline questions — a key purchase factor
Implement get_shipping_options with address + cart parameters
4
Checkout uses browser JavaScript
Agents cannot complete purchases programmatically — transactions stop at the final step
Add checkout tool with requestUserInteraction() for payment security
The mechanics are straightforward: stores with structured WebMCP tool definitions give agents typed responses instead of guesswork. That reliability compounds with every transaction.
10 / Implementation
Getting Started
Three paths to agent readiness. Choose the one that matches your tech stack and timeline.
CMS Plugin
Shopify, WooCommerce, Magento, BigCommerce
5–20 min
estimated effort
1
Install Plugin
5 minOne-click install from your platform's app/extension store.
2
AI Co-Pilot Scan
2 minCo-PilotAutomatically discovers your product catalog, forms, checkout flow.
3
Review Tools
10 minCo-PilotReview 8 auto-generated tool definitions. Customize descriptions. Toggle on/off.
4
Go Live
1 clickDeploy. Monitor agent traffic in real-time dashboard.
Not sure which path? Start with a free Quick Score — it analyzes your site and recommends the best implementation path.
Or jump straight in with our E-Commerce WebMCP Template →
11 / Pricing
Pricing That Scales With Agent Traffic
Free Quick Score for any store. Paid plans unlock the Co-Pilot, analytics, and competitive intelligence.
MonthlyAnnual-20%
No transaction fees
14-day free trial
Cancel anytime
Free
Quick Score + basic tool analysis for any store.
$0/mo
- Quick Agent Readiness Score
- Basic tool detection (3 pages)
- Improvement recommendations
- Community support
Starter
For small stores getting agent-ready.
$39/mo
Save $120 /year
- Deep Score (full site scan)
- Up to 8 tool definitions
- AI Co-Pilot tool generation
- Basic agent analytics
- Email support
Most Popular
Pro
For growing stores optimizing agent revenue.
$119/mo
Save $360 /year
- Everything in Starter
- Unlimited tool definitions
- Advanced agent analytics
- Competitive intelligence
- A/B test tool descriptions
- Priority support
- Custom tool schemas
Business
For multi-store operations and agencies.
$319/mo
Save $960 /year
- Everything in Pro
- Up to 10 stores
- Multi-store dashboard
- White-label agent reports
- API access
- Dedicated account manager
- Custom integrations
Enterprise
For large retailers with custom needs.
Custom
- Everything in Business
- Unlimited stores
- Custom SLA
- On-premise deployment option
- Dedicated engineering support
- Custom AI Co-Pilot training
- Multi-region, multi-currency
- SOC 2 compliance
CMS Plugin Add-Ons
Shopify
From $29/mo
Included in Starter+ plans
WooCommerce
From $29/mo
WordPress plugin
Magento
From $99/mo
Adobe Commerce compatible
BigCommerce
From $29/mo
Native integration
Common Questions
WebMCP is an open protocol — like HTTP. But implementing it well requires tooling. WebMCP provides the AI Co-Pilot that generates tool definitions, the analytics that track agent interactions, and the competitive intelligence that optimizes your store. You could implement WebMCP yourself (it's open!), but WebMCP gets you there in 5 minutes instead of weeks.
It's backed by Google and Microsoft and already shipping in Chrome 146. But even in the worst case, the structured data you create improves your existing SEO and accessibility. There's no downside scenario.
Yes. The Free tier gives you a Quick Score and basic tool analysis with no signup required. Paid plans include a 14-day free trial with full access.
ROI depends on how quickly agent traffic grows for your category. Once live, your store becomes discoverable by AI agents in Chrome. The timeline for meaningful agent-driven revenue will vary by product category, traffic volume, and agent adoption rates — which are still early.
Not for CMS platforms (Shopify, WooCommerce, etc.) — the plugin handles everything. For custom/headless builds, you'll need a developer familiar with your stack. The AI Co-Pilot can generate initial tool definitions to accelerate setup.
12 / The Close
WebMCP Is Live in Chrome 146.
See How Your Store Compares.
Google released WebMCP recently. It's available in Chrome 146 early preview. AI agents can already call structured tools on websites that implement the protocol.
As AI agents become a meaningful traffic source, the quality of your WebMCP tool definitions could matter as much as your product descriptions. Structured data for agents follows the same logic as structured data for search engines.
The protocol is open, the spec is public, and Chrome support is already shipping. Explore what it means for your store.
Takes 60 seconds. No credit card. No signup for Quick Score.
Built on the official WebMCP spec by Google & Microsoft
No credit card required
Cancel anytime