io.github.bluestratusPublished by GitHub account bluestratusv0.2.0Updated Oct 24, 2026
Match shopper intent to a merchant's products and fetch live price, stock and product Q&A.
A remote MCP server that lets an assistant match a shopper's intent to a merchant's products and fetch live price, stock and product detail.
Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.
Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.
Read https://sourceweft.com/mcp/io.github.bluestratus%2Fagenticfeed/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.Add this to your client's mcpServers config.
{
"mcpServers": {
"agenticfeed": {
"type": "http",
"url": "https://agenticfeed.ai/mcp"
}
}
}[Version] [License: MIT] [Reference implementation]
An open specification for making ecommerce product catalogues readable by AI shopping agents.
AI assistants like ChatGPT, Claude, Perplexity, and Gemini are already recommending products to millions of shoppers. They do not browse category pages or read banner ads. They consume structured data and reason about which products best match the buyer's intent. Most merchant websites are invisible to them.
This standard defines a lightweight, discoverable format that gives AI agents exactly what they need.
Add one line to the <head> of every page on your website:
That tag tells any AI agent or crawler where your structured product feed lives. The rest of this document describes what that feed should contain.
An agentic feed is a structured product data feed built for AI agents rather than search engine crawlers or human browsers.
A traditional product page is designed to be rendered and read by a person. A Google Merchant Center feed is designed to be parsed by a price comparison engine. An agentic feed is designed to be reasoned about by an AI.
The key difference is intent data. Where a merchant feed tells an agent "here is a cordless drill, it costs £29.99 and it is in stock," an agentic feed tells it "here is a cordless drill that answers the question what drill do I need for assembling flat-pack furniture, solves the problem I keep stripping screws with my old drill, and fits the use case home DIY for a first-time homeowner."
That is the layer of context an AI agent needs to make a confident recommendation to a specific buyer.
AI shopping is already happening. The tooling to serve it well does not yet exist as an open, portable standard.
Search engines standardised web content discovery through sitemaps, robots.txt, and canonical tags. RSS standardised content syndication through a single autodiscovery tag. Neither was designed for the kind of structured reasoning that AI agents perform when they decide what to recommend.
This specification fills that gap. It defines:
The format is intentionally minimal. It builds on schema.org types that crawlers already understand. It adds the intent layer that makes AI recommendation possible.
A Google Merchant Center feed tells a machine what a product is. An agentic feed tells it why a specific buyer should choose it.
The two are complementary. Many merchants use a GMC feed as the data source for generating an agentic feed.
The discovery mechanism follows the same autodiscovery pattern the web has used for decades:
When an AI agent or crawler visits a merchant's website, it reads the page <head>. If it finds a rel="agenticfeed" tag, it knows exactly where to fetch structured product data without being told the URL in advance.
This means:
Step 1. Add the discovery tag to every page <head>:
Step 2. Serve a JSON-LD feed index at that URL (see examples/feed.json):
Step 3. Serve intent endpoints for each category (see examples/questions.json).
Step 4. Serve product detail documents for each product (see examples/product.json).
For Shopify merchants, agenticfeed.ai handles all of this automatically including injecting the discovery tag into your theme.
Place this tag in the <head> of every page. It must appear in the server-rendered HTML, not injected by JavaScript, so crawlers can find it without executing scripts.
Content-Type: application/ld+json
Schema.org type: DataFeed
The feed index is the entry point for any agent reading your feed. It identifies the merchant, lists intent endpoints, and provides the URL template for resolving individual products.
See examples/feed.json for a complete example.
Agent workflow:
<head>dataFeedElement to find intent endpointsagenticfeed.resolution.product.url templateIntent endpoints are JSON-LD ItemList documents grouped by product category. There are three intent types:
Questions — natural language questions a buyer asks before purchasing.
Schema.org type: Question with suggestedAnswer pointing to the product URL.
See examples/questions.json
Problems — pain points or needs the product addresses.
Schema.org type: ListItem with name (the problem) and url (the product).
See examples/problems.json
Use cases — specific scenarios or goals the product fits.
Schema.org type: ListItem with name (the use case) and url (the product).
See examples/use-cases.json
Content-Type: application/ld+json
Schema.org type: Product
Each product has its own document at a stable URL. It combines standard schema.org Product data with the intent content under an agenticfeed namespace.
See examples/product.json for a complete example.
All product URLs in an agentic feed should carry UTM parameters so merchants can measure AI agent traffic in their analytics:
This lets merchants filter and report on sessions and orders that originated from an AI agent recommendation in Google Analytics, Shopify Analytics, or any platform that reads UTM parameters.
The discovery tag in the merchant's <head> serves as proof of domain ownership before a feed goes live. A validator fetches the merchant's homepage, parses the <head>, and confirms:
rel="agenticfeed" tag is presenthref matches the expected feed URL for that merchantThis prevents one merchant from claiming another merchant's domain in a feed.
The passive feed endpoints let AI agents crawl product data at their own pace. The Agent Query API is the active layer — an AI agent or application can send a natural language buyer intent and receive back a ranked list of matched products with reasons.
This is agent-to-agent communication. Instead of a human typing into a search box, one AI asks another AI to find the best matching products.
Endpoint
One API key is required:
Agenticfeed API key — passed as a Bearer token in the Authorization header. This identifies who is making the request and controls access to the feed data. Generate one in Dashboard › API Keys for any paid plan (Growth and above). No approval needed — keys are available instantly.
Returns a schema.org ItemList of matched Product entries, each with a match_reason explaining why that product fits the buyer's need.
Content-Type: application/ld+json
The match_reason field is designed to be passed directly to the end buyer by the calling agent. It explains the recommendation in plain language without requiring the agent to do any additional reasoning.
The reference implementation publishes an A2A-compatible Agent Card at:
This file describes the agent's capabilities, accepted inputs, output format, and authentication requirements in a machine-readable format. AI platforms that implement Google's Agent-to-Agent (A2A) protocol can discover and call the query endpoint automatically using this card.
All examples are in the examples/ directory:
agenticfeed.ai is the hosted reference implementation of this standard.
It provides:
/agent/query for natural language product matching (included with paid plans)/.well-known/agent.jsonMerchants who use agenticfeed.ai get a fully conformant agentic feed without writing any code. Developers building AI agents can query product data by buyer intent using the Agent Query API, included with all paid merchant plans.
Issues and pull requests are welcome at github.com/bluestratus/agenticfeed.
The goal of this standard is to be minimal and stable. Proposals that add complexity without clear benefit to AI agents or merchants will not be merged. The best contributions are real-world implementation experience, edge cases, and corrections.
Published under the MIT Licence. Anyone is free to implement compatible agentic feeds using this standard without restriction.
Source: README.md at commit 914baba
Resolve a merchant's website domain to their Agenticfeed merchant ID and confirm they publish an Agentic Feed. Call this first when you only know a store's domain; use the returned merchant_id with get_product or the discovery feeds.
domainrequiredGet full detail for a single product by merchant_id and product_id: name, price, currency, availability, image, and the buyer questions, problems, and use cases it maps to. Returns a schema.org Product.
merchantrequiredproduct_idrequiredList a merchant's in-stock products (product_id, title, price, availability, url). Call after find_merchant to see the catalogue, then get_product for full detail. For a specific shopper need, prefer search_products.
limitmerchantrequiredFind a merchant's products that match a shopper's need, question, problem, or use case. Pass the buyer intent in natural language (e.g. "a light strategy game for 4 players under thirty pounds"); returns ranked in-stock products with the reason each fits, drawn from Agenticfeed's intent mapping (questions, problems, use cases). Call find_merchant first to get the merchant ID.
limitqueryrequiredmerchantrequired100% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
<link rel="agenticfeed" type="application/json" href="https://yourdomain.com/feed.json">rel="stylesheet" tells browsers where to find CSSrel="icon" tells browsers where to find the faviconrel="alternate" tells crawlers where to find RSS feedsrel="agenticfeed" tells AI agents where to find structured product data<link rel="agenticfeed" type="application/json" href="https://yourdomain.com/feed.json">{ "@context": "https://schema.org", "@type": "DataFeed", "name": "Your Store Name", "url": "https://yourdomain.com/feed.json", "provider": { "@type": "Organization", "name": "Agenticfeed", "url": "https://agenticfeed.ai" }, "dataFeedElement": [ { "@type": "DataFeedItem", "name": "Questions", "url": "https://yourdomain.com/questions.json" }, { "@type": "DataFeedItem", "name": "Problems", "url": "https://yourdomain.com/problems.json" }, { "@type": "DataFeedItem", "name": "Use Cases", "url": "https://yourdomain.com/use-cases.json" } ]}<link rel="agenticfeed" type="application/json" href="{absolute-url-to-feed-index}">utm_source=agenticfeed&utm_medium=ai-agentPOST https://agenticfeed.ai/agent/queryContent-Type: application/jsonAuthorization: Bearer af_your_api_key{ "query": "ergonomic chair for someone with lower back pain, works from home 10 hours a day, small home office, budget under £300", "customer_guid": "a9a8378c94", "limit": 5}{ "@context": "https://schema.org", "@type": "ItemList", "query": "ergonomic chair for back pain...", "numberOfItems": 3, "itemListElement": [ { "@type": "Product", "position": 1, "name": "HM Seating Contessa Chair", "url": "https://merchant.com/contessa?utm_source=agenticfeed&utm_medium=ai-agent", "image": "https://merchant.com/images/contessa.jpg", "category": "office chairs", "offers": { "@type": "Offer", "price": "249.00", "priceCurrency": "GBP", "availability": "https://schema.org/InStock" }, "agenticfeed": { "product_id": "b3f1e29a4c", "match_reason": "Matches adjustable lumbar support, compact footprint for small spaces, and falls within the stated budget." } } ]}GET https://agenticfeed.ai/.well-known/agent.json