An agent card is a structured JSON file published on your website that tells AI agents what your business does, how it operates, and what actions they are allowed to take. For ecommerce stores specifically, it is the file that tells AI shopping agents what you sell, what your shipping and return policies are, how to browse your catalog, and whether they can initiate a checkout. Without it, an AI agent trying to work with your store has to guess. Guessing leads to incomplete recommendations, missed interactions, and lost visibility in the AI commerce channels that are growing fastest in 2026.
This guide explains what an agent card is in plain language for ecommerce store owners, the difference between informational and action-oriented agent cards, what a realistic ecommerce agent card actually looks like, and how to publish one on your store today.
Is your store ready for AI agent commerce?
We build the agentic commerce infrastructure — agent cards, schema, feeds, and buying guide content — that makes ecommerce stores visible and actionable to AI agents.
The Quick Take: What an Agent Card Does vs. What Your Other Files Do
| File | What It Does for Your Store |
|---|---|
| robots.txt | Controls which AI crawlers can access your store |
| Schema markup | Tells AI what your products, prices, and reviews mean |
| llms.txt | Gives AI a map of your most important pages |
| Product feed | Syndicates your catalog to shopping channels |
| Agent card | Tells AI agents how to interact with your store |
The Takeaway: A product feed tells AI what you sell. An agent card tells AI how to buy it. They are not the same thing and you need both.
💡 Pro Tip: Before reading further, visit yourdomain.com/.well-known/agent-card.json in your browser. If you see a 404 error or a blank page, your store does not have an agent card yet. You are not alone — almost no ecommerce stores have published one. That gap is exactly where the early mover advantage lives right now.
Table of Contents
→ The AI Shopping Scenario You Need to Understand
→ What an Agent Card Is and What It Does for Your Store
→ Informational vs. Action-Oriented: The Two Types of Ecommerce Agent Cards
→ How an Agent Card Differs From Your Feed, Schema, and llms.txt
→ What an Ecommerce Agent Card Actually Looks Like
→ Where to Publish It and How to Reference It
→ Why Ecommerce Stores Should Care Now
→ The Bottom Line on Agent Cards for Ecommerce
→ FAQ: Common Questions About Agent Cards for Ecommerce
The AI Shopping Scenario You Need to Understand
A shopper asks ChatGPT to find trail running shoes under $150, size 11, with free shipping to California. Your store sells the right product at the right price. But it does not show up in the recommendation. Not because your product is wrong. Because the AI agent could not reliably interpret your inventory data, shipping rules, or return policy from your product pages alone.
AI agents making shopping recommendations in 2026 are not just reading your product listings. They are trying to understand whether your store is a reliable, structured endpoint they can confidently recommend to a shopper. Stores that give agents clean, structured signals get recommended. Stores that make agents guess get skipped.
That is the agent card problem. And it is more common than almost any ecommerce guide has documented yet, because agent cards are new enough that most store owners have never heard of them.
The good news: an agent card is not a complex technical project. It is a structured description of your store written for machines. If you know your products, policies, and checkout flow, you already have most of what you need to write one.
💡 Pro Tip: Open ChatGPT right now and ask it to recommend a store in your product category with your specific constraints — price range, shipping region, use case. Note which stores appear and which do not. The stores that appear consistently are giving AI agents cleaner structured signals. The agent card is one of the most direct ways to improve the quality of those signals.
What an Agent Card Is and What It Does for Your Store
An agent card is a JSON file published at a standardized path on your domain that tells AI agents what your store sells, how it operates, and what they are allowed to do. The official A2A (Agent-to-Agent) protocol specification, developed by Google and supported by over 150 organizations as of 2026, defines the agent card as the foundational discovery document for any business that wants AI agents to interact with it reliably.
For an ecommerce store, an agent card should cover:
- Store name and primary product categories
- Catalog or feed location
- Shipping regions and delivery expectations
- Return policy summary
- Supported actions such as browse, add to cart, checkout, or order lookup
- Constraints and fallback paths
- Contact and booking endpoints
The best analogy for an ecommerce store owner: an agent card is the briefing document you would hand a personal shopping assistant before sending them to buy something for a customer. Without that briefing, the assistant has to guess what your policies are, what you stock, and what they are allowed to do. Guessing leads to errors, abandoned sessions, and recommendations that route shoppers to your better-documented competitors instead.
The A2A protocol specifies that agent cards should be published at yourdomain.com/.well-known/agent-card.json, following RFC 8615 standards for well-known URI discovery. This path is where AI agents look first when evaluating whether a domain has defined its agent interaction parameters.
đź’ˇ Pro Tip: The agent card specification is part of the A2A (Agent-to-Agent) open protocol announced by Google in April 2025 and now backed by over 150 organizations. It is an emerging standard, not a locked-down requirement. Early adoption is the advantage. The cost of publishing an agent card is low. The cost of being undiscoverable to AI agents as agentic commerce grows is not.
Informational vs. Action-Oriented: The Two Types of Ecommerce Agent Cards
Ecommerce agent cards fall into two categories, and most stores should start with the first before building toward the second.
The Informational Agent Card
An informational agent card helps AI models accurately describe, cite, and recommend your store. It does not enable transactions, but it significantly improves discovery accuracy and citation quality.
Example use case: a buyer asks Perplexity to compare sustainable activewear brands under $100 with free returns. An informational agent card ensures your store is described accurately, including your product categories, price range, shipping regions, return policy, and brand positioning. Without it, the AI has to infer these details from your website content, which leads to incomplete or inaccurate descriptions.
Who needs it: every ecommerce store, regardless of size or platform. This is the baseline.
The Action-Oriented Agent Card
An action-oriented agent card defines what AI agents can actually do on your store, including browsing the catalog, checking inventory, creating a cart, starting checkout, looking up order status, or routing inquiries to a human.
Example use case: a buyer uses an AI shopping assistant to purchase a birthday gift with specific constraints. The agent needs to know whether your store supports guest checkout, whether discount codes can be applied programmatically, whether inventory is real-time, and what happens if the agent encounters an error. If those parameters are not defined, the agent may fail silently or route the buyer to a store that has defined them.
Who needs it: stores preparing for AI-initiated transactions and agentic commerce workflows.
| Agent Card Type | Primary Function | Who Needs It Now |
|---|---|---|
| Informational | Improves discovery accuracy and citation quality | Every ecommerce store |
| Action-oriented | Enables AI agents to take defined actions on your store | Stores ready for agentic commerce transactions |
The practical recommendation: publish an informational agent card this week. Build toward action-oriented capabilities as agentic commerce adoption grows in your category. The informational card has zero technical prerequisites and takes under two hours to write and publish.
How an Agent Card Differs From Your Feed, Schema, and llms.txt
The most common confusion about agent cards is treating them as redundant with files ecommerce stores already have. They are not redundant. Each file serves a different layer of AI interaction, and none of them replaces the others.
| File | Layer | What It Covers | What It Does Not Cover |
|---|---|---|---|
| robots.txt | Access | Which crawlers can enter your store | What they find when they get there |
| Schema markup | Understanding | What individual pages and products mean | How agents should interact with the store |
| Product feed | Syndication | What your catalog contains and at what price | Policies, constraints, and interaction parameters |
| llms.txt | Discovery | Which pages matter and what the brand is about | Actions agents can take or workflows they can follow |
| Agent card | Interaction | What agents can do and how to do it reliably | Individual product data or page-level content |
The key distinction that matters most for ecommerce: a product feed tells AI what you sell. An agent card tells AI how to buy it. They operate at different layers of the commerce stack and a store needs both to be fully AI-ready.
For the full picture of all four AI infrastructure files and how they work together, see: The 4 Files Every Website Needs for AI Discovery.
What an Ecommerce Agent Card Actually Looks Like
The most useful thing this guide can do is show you what an actual ecommerce agent card looks like, not just describe it abstractly. Here are two realistic examples for a DTC activewear store: one informational and one action-oriented.
Example 1: Informational Agent Card
This version helps AI models accurately describe and recommend the store. It requires no technical integration and can be published immediately.
{
"schemaVersion": "1.0",
"version": "1.0",
"name": "ProGrip Athletics",
"url": "https://progripathletics.com",
"type": "ecommerce_store",
"description": "ProGrip Athletics sells premium yoga and activewear for daily practitioners. Products include yoga mats, resistance bands, and performance apparel. All products ship free to the US and Canada with a 60-day return policy.",
"audience": "Yoga practitioners, fitness enthusiasts, and active lifestyle consumers seeking durable, performance-focused gear.",
"productCategories": [
"Yoga mats",
"Resistance bands",
"Performance apparel",
"Gym accessories"
],
"priceRange": {
"min": 24,
"max": 149,
"currency": "USD"
},
"shipping": {
"regions": ["US", "CA"],
"freeShippingThreshold": 50,
"estimatedDelivery": "3-5 business days"
},
"returns": {
"windowDays": 60,
"condition": "Unused with original packaging"
},
"catalogUrl": "https://progripathletics.com/collections/all",
"preferred_citation": "ProGrip Athletics (progripathletics.com)",
"do_not_represent_as": [
"A luxury brand",
"A brick-and-mortar retailer",
"A subscription service"
]
}
Example 2: Action-Oriented Agent Card
This version adds the skills block that defines what AI agents are allowed to do on the store. It is built on top of the informational card above.
{
"schemaVersion": "1.0",
"version": "1.0",
"name": "ProGrip Athletics",
"url": "https://progripathletics.com",
"type": "ecommerce_store",
"description": "ProGrip Athletics sells premium yoga and activewear for daily practitioners.",
"capabilities": {
"streaming": false,
"pushNotifications": false,
"agentInteractionsEnabled": true,
"authenticationRequired": false
},
"skills": [
{
"name": "Browse Catalog",
"description": "Direct agents to the full product catalog or a specific category.",
"action": "redirect",
"endpoint": "https://progripathletics.com/collections/all",
"triggerIntents": [
"show me products",
"browse catalog",
"what do you sell",
"find a yoga mat"
]
},
{
"name": "Check Product Availability",
"description": "Route agents to a specific product page for real-time inventory.",
"action": "redirect",
"endpoint": "https://progripathletics.com/collections/all",
"triggerIntents": [
"is this in stock",
"check availability",
"do you have size"
]
},
{
"name": "Start Checkout",
"description": "Route agents to the cart or checkout flow.",
"action": "redirect",
"endpoint": "https://progripathletics.com/cart",
"triggerIntents": [
"add to cart",
"buy now",
"purchase",
"checkout"
]
},
{
"name": "Contact Support",
"description": "Route agents to customer support for order or policy questions.",
"action": "redirect",
"endpoint": "https://progripathletics.com/pages/contact",
"triggerIntents": [
"track my order",
"return policy",
"shipping question",
"contact support"
]
}
],
"constraints": {
"doNotMakePricingCommitments": true,
"doNotGuaranteeDeliveryDates": true,
"checkoutRequiresHumanCompletion": true
},
"fallback": {
"description": "For any request not covered above, route to the contact page.",
"endpoint": "https://progripathletics.com/pages/contact"
}
}
Every field in both examples maps to a real store attribute you already know. Store name, product categories, price range, shipping rules, return policy, and contact endpoint. The thinking is harder than the file. Once you know what to write, the JSON structure itself is straightforward.
💡 Pro Tip: Start with the informational version. Fill in each field with accurate information about your store. Keep descriptions factual and specific — not marketing language. AI agents evaluate agent cards for accuracy and specificity. A description that says “premium quality products” tells an agent nothing. A description that says “yoga mats in 4mm, 6mm, and 8mm thickness, priced $49 to $149, free shipping to US and Canada” tells an agent exactly what it needs to make a confident recommendation.
Where to Publish It and How to Reference It
Publishing an agent card has two steps: creating the file and placing it at the correct path on your server.
The Correct Path
The A2A protocol specification defines the canonical path as yourdomain.com/.well-known/agent-card.json. This follows RFC 8615 standards for well-known URI discovery, which is how AI agents know where to look for the file without being told explicitly. Publishing at any other path means agents may not find it automatically.
How to Upload It
For WordPress and WooCommerce sites: connect to your server via FTP or cPanel File Manager, navigate to your WordPress root directory, create a folder named .well-known if it does not already exist, and upload your agent-card.json file into that folder. Test by visiting yourdomain.com/.well-known/agent-card.json in your browser. You should see the raw JSON.
For Shopify stores: Shopify does not allow direct file uploads to the root directory. The current workaround is to publish the agent card content as a page at a consistent URL and reference that URL, or work with a developer to serve it via a custom route. This limitation is a known gap in Shopify’s current infrastructure that is expected to be addressed as agent card adoption grows.
How to Reference It
Once published, reference your agent card in your llms.txt file if you have one. Add a line like: Agent Card: https://yourdomain.com/.well-known/agent-card.json. This tells AI language models that an agent card exists and where to find it, which improves the probability that agents checking your llms.txt will also read your agent card.
Keep It Updated
An agent card is only as useful as it is accurate. Update it when your shipping regions change, when you add major product categories, when your return policy changes, or when you add new capabilities. An outdated agent card is worse than no agent card for accuracy-sensitive recommendations.
💡 Pro Tip: WP Engine users should note that the .well-known directory may require an additional access rule in your .htaccess file if WP Engine’s security layer blocks direct access to it. Add this rule to allow public access: <Files "agent-card.json"> Allow from all </Files>. Test by visiting the URL in an incognito window after upload to confirm it is publicly accessible.
Why Ecommerce Stores Should Care Now
Agent cards matter now because AI agents are already making purchase recommendations and some are beginning to support cart and checkout flows. The stores that have defined how agents should interact with them will be easier to recommend, easier to trust, and easier to transact with as agentic commerce grows.
The data makes the trajectory clear. During Shopify’s Q1 2026 earnings call, president Harley Finkelstein reported that AI-driven traffic to Shopify stores had grown 8x year over year, while orders from AI-powered searches had increased nearly 13x. That growth is happening now, not in some future state. The stores capturing it are the ones that have done the infrastructure work: schema, feeds, crawler access, and increasingly, agent cards.
The competitive window is real. Neither Shopify nor WooCommerce publishes an agent card by default. No platform generates one automatically. Every agent card currently live on an ecommerce store was put there deliberately by someone who understood why it matters. That is a very small number of stores right now. The gap between stores that have one and stores that do not will narrow as adoption grows. The stores that publish now build familiarity with AI systems before the channel becomes crowded.
For the complete ecommerce AI visibility picture, see our guide to AI search visibility for ecommerce brands. For the full four-file infrastructure framework that the agent card fits into, see: The 4 Files Every Website Needs for AI Discovery. For WooCommerce-specific schema implementation that works alongside your agent card, see: WooCommerce Product Schema: The Complete Setup Guide. For the platform-agnostic schema guide, see: Product Schema for Agentic Commerce.
The Bottom Line on Agent Cards for Ecommerce
An agent card is the file that tells AI agents how to work with your store. Your robots.txt controls who can enter. Your schema explains what your products mean. Your product feed syndicates your catalog. Your agent card defines the interaction layer: what agents are allowed to do, how your store operates, and what they should tell shoppers about you.
Most ecommerce stores have one or two of these files in place. Almost none have the agent card. That gap is not permanent, but it is currently an advantage for any store that closes it. An informational agent card for a DTC store takes under two hours to write and a few minutes to upload. The technical barrier is low. The strategic upside is real and growing.
If you want help building the complete agentic commerce infrastructure for your store — agent card, schema stack, Merchant Center feed, and buying guide content — that is exactly what our Agentic Commerce service covers.
🎯 Get Your Ecommerce Store Agent-Ready
We build the complete agentic commerce infrastructure — agent card, schema stack, Merchant Center feed, and buying guide content — that makes your store visible, accurate, and actionable to AI agents. Done and Indexed starts at $2,500.
→ Book Your Free Strategy Call
Almost no ecommerce stores have a live agent card. Publishing one today puts you ahead of the vast majority of your competitors in AI agent discovery.
Frequently Asked Questions About Agent Cards for Ecommerce
What is an agent card for an ecommerce store?
An agent card is a structured JSON file published at yourdomain.com/.well-known/agent-card.json that tells AI agents what your store sells, how it operates, and what actions they are allowed to take. It is the file that defines the interaction layer between your store and AI shopping agents, covering product categories, shipping policies, return policies, supported actions, and fallback instructions.
Do Shopify stores need an agent card?
Yes. Shopify does not generate an agent card automatically. While Shopify merchants benefit from Agentic Storefronts for catalog syndication, an agent card defines the interaction parameters and constraints that catalog syndication does not cover. Shopify’s current infrastructure has a limitation in that it does not support direct file uploads to the root directory, so publishing an agent card may require a developer workaround until Shopify addresses this gap natively.
Does WooCommerce support agent cards?
WooCommerce does not generate agent cards automatically, but WooCommerce stores can publish them more easily than Shopify stores because WordPress allows direct file uploads to the root directory via FTP or cPanel. Create a .well-known folder in your WordPress root, upload your agent-card.json file, and test by visiting yourdomain.com/.well-known/agent-card.json in your browser.
Where do I publish an agent card?
The A2A protocol specification defines the canonical path as yourdomain.com/.well-known/agent-card.json, following RFC 8615 standards for well-known URI discovery. For WordPress and WooCommerce sites, upload the file via FTP or cPanel File Manager to the .well-known folder in your WordPress root directory. Reference it in your llms.txt file once published.
Is there an official agent card standard?
The agent card is defined by the A2A (Agent-to-Agent) protocol, an open standard announced by Google in April 2025 and supported by over 150 organizations as of 2026. It is a real, actively maintained protocol with a published specification at a2a-protocol.org. The spec is still evolving toward version 1.0, but the core agent card structure is stable enough for production use. Early adoption is the advantage while the ecosystem is still forming.
How is an agent card different from a product feed?
A product feed tells AI what you sell, including your catalog, prices, availability, and product attributes. An agent card tells AI how to buy it, covering your policies, your supported actions, your constraints, and your fallback paths. They operate at different layers of the commerce stack and a fully AI-ready ecommerce store needs both.
How long does it take to create an agent card for an ecommerce store?
An informational agent card for a standard DTC ecommerce store takes under two hours to write and a few minutes to upload. The thinking is harder than the file. You need to know your product categories, price range, shipping regions, return policy, and contact endpoints. Once you have that information gathered, the JSON structure itself is straightforward. An action-oriented agent card with defined skills and workflows takes longer and may benefit from developer input.
What is the difference between an informational and action-oriented agent card?
An informational agent card helps AI models accurately describe and recommend your store. It improves discovery and citation quality without enabling transactions. An action-oriented agent card defines specific workflows AI agents can execute on your store, such as browsing catalog, checking inventory, initiating checkout, or routing inquiries. Most ecommerce stores should publish an informational card first and build toward action-oriented capabilities as agentic commerce adoption grows.