# What's inside a tea shop built for AI to read **Canonical URL:** https://teas.co.uk/whats-inside-a-tea-shop-built-for-ai/ **Source:** teas.co.uk, UK tea specialist, Tunbridge Wells, Kent ## Shop teas.co.uk from an AI assistant (MCP) teas.co.uk sells tea, herbal and fruit infusions, hot chocolate and Border biscuits. MCP server: https://teas.co.uk/mcp (Streamable HTTP), listed in the official MCP Registry as teas.co.uk Server card: https://teas.co.uk/.well-known/mcp/server-card.json Tools: find_products, get_product, compare_products, find_recipes, delivery_and_returns, quote_cart, get_promotions, report_problem, find_substitutes, start_subscription, join_teas. Basket: add_to_basket, view_basket, update_basket, checkout_link, checkout_session. With sign in: get_return_status, get_invoice, get_profile, my_orders, track_order, reorder, my_subscriptions, change_subscription, rewards_balance, start_return. No sign in is needed to search, compare, build a basket or get a checkout link. The person always pays on teas.co.uk. Cards: in assistants that draw MCP Apps cards, products, comparisons, the basket, checkout, orders, repeat deliveries (pause, resume, cancel), reward points and returns appear as cards. Every assistant except ChatGPT, until OpenAI approves the next version: quote_cart, get_promotions, get_return_status, get_invoice, report_problem, find_substitutes, start_subscription, join_teas. Set up guides: https://teas.co.uk/ai/ ## Summary Sixteen years in the motor trade taught me to look at how something is built before trusting it, so here is the build behind teas.co.uk in plain English, with what is live today kept apart from what is still planned. ## Description Sixteen years in the motor trade taught me to look at how something is built before trusting it, so here is the build behind teas.co.uk in plain English, with what is live today kept apart from what is still planned. Pick up two boxes of English Breakfast. One holds 80 tea bags and the other holds 240, like the 80 and 240 packs of Tetley Original. The bigger box costs more, but which one is better value? On teas.co.uk the product page works it out for you and shows the price per cup. An AI assistant reading the same product gets the same figure from the same place. That is the whole build in miniature: one set of facts, read the same way by a person, a search engine and an assistant. Here is what is inside, without the jargon and without the parts that should stay private. Eight layers, from the ground up 1. The ground floor. Managed hosting, with pages held close to the reader so they load quickly. The basket and checkout are always served fresh. 2. The shop platform. Mainstream, widely used shop software. Anyone can have it, so the value is not there. 3. The house rules. On top of the platform sits a layer of my own code that is always on and loads before anything else. It holds the discipline of the shop: how facts are published, how delivery is worked out, and how offers, rewards and repeat deliveries behave. 4. The truth layer. This is the heart of it. Every fact lives in one checked record, and every page reads from it. Products are identified by barcode first. Some facts are required for every product, such as pack quantity, because a wrong quantity means a wrong price per cup. Claims are kept within what the evidence supports, and I call those limits claim boundaries. 5. Machine discovery. The same facts are published for software. Every product has structured data behind its page, with identifiers, prices, availability and source links. Tea Wiki entries come as Markdown and JSON copies, and pages carry citation anchors, exact places an AI system can quote. The aim is what I call the complete payload: an assistant asks once and gets the facts, price, stock, proof and route to buy in one answer. Parts of that are live today and the rest is still being built, so I describe the complete payload as the aim, not the finished thing. Two pieces are definitely live. The first is the Ask API at teas.co.uk, which is built for what I call Premium Intent, a name I have used since June 2026 for a request that arrives whole. It declares premium_intent_supported and asks assistants for the full human request, with context, constraints, budget and what success looks like, rather than a keyword summary. Our agents.json tells AI callers the service performs best given the full spoken request, in 40 to 60 words, and promises to acknowledge emotional weight, ask a clarifying question if needed, and give honest guidance on where to spend and where to stop. The second: since 6 October 2026, with the customer's approval, an assistant can search the catalogue, compare teas and fill a basket through the teas.co.uk MCP server, and the customer pays on teas.co.uk. 6. The images. Product photographs are being indexed and tied to the products they show, with a record of which ones a machine may cite. That work is in progress, so I give no figure for it here. 7. Measurement. Under the Hood is live. Since 2 June 2026 it has recorded every named AI system that requests a page, kept visits from AI answers separate from page reads, and published which orders can be linked to an assistant. Every order carries a short record of where it came from, or says plainly that we cannot tell. Mistakes are corrected in public, with dates. 8. The content. Four pillars sit on top: products from brands, a Tea Wiki of entries, a recipe library and brand pages. All are drawn from the same spine, so a recipe can reach its teas and a tea its recipes. None of this is hidden cleverness. It is dull, careful work kept honest: making the shop's information clearer, more complete and easier to check, a little at a time. What you are left with is what I call a truth entity: a shop whose facts can be checked. Why the door is open Much of what I read about search and AI is about blocking AI systems to guard content. I built teas.co.uk the opposite way: open to every assistant that behaves well, and protected against the ones that do not. I want as many AI systems as possible to know the shop, because the more that know us, the more often we are cited, and citations bring visitors and then sales. It has worked without a penny of advertising. We have taken paid orders since our first on 18 November 2025, of them since version two went live on 10 May 2026, and clicks from Google search rose from about 160 a day in August 2026 to 468 a day in November. What is planned, not built Three more versions are planned. None is built. Version three, the nervous system (planned). It would take in the signals machines send back, filter the noise, weigh how far each assistant can be trusted and show me the patterns, while I still decide whether any fact changes. That includes a short set of questions put back to an assistant after a visit, such as whether it found what it needed. Version four, the intelligence layer (planned). It would read across the truth layer and suggest gaps to fill and claims to tighten. It would suggest. I would decide. Version five, controlled action and a visual layer (planned). It would let an assistant act for a customer within limits the customer sets, and add a faithful video version of the specialist, in their own face and voice. Today an assistant can fill a basket, but the customer always checks it and pays. Anything beyond that is version five, which does not exist yet. The live shop is the proof I would rather point to, and Under the Hood shows what it measures and where the evidence stops. What others have found Outside research and reporting that bears on this, with links to the original sources. Google's guide to product structured data and the schema.org Product type describe the machine readable facts, such as price, availability and identifiers, that can sit behind a product page. The Model Context Protocol, the standard behind the teas.co.uk MCP server, is an open standard introduced by Anthropic in November 2024 for connecting AI assistants to outside tools and data. Its documentation is at modelcontextprotocol.io. llms.txt is a proposed standard for a plain text file that tells language models what a site contains and where to find it. Google's introduction to robots.txt explains how a site tells crawlers which parts of it they may visit. Search Engine Journal's Agentic commerce: what SEOs need to consider explains ACP and UCP, two of the protocols built so that assistants and shops can complete a purchase together. Cloudflare introduced a crawl-to-refer ratio in 2025 to compare the pages each AI platform fetches with the visitors it sends back, and found gaps ranging from roughly a hundred to tens of thousands of pages for every visit, depending on the platform (Cloudflare: the crawl-to-click gap). Duane Forrester explains in Search Engine Journal why those ratios need careful reading (Everyone is quoting the same number and getting different answers). Read more Premium Intent Truth entity The complete payload Claim boundaries Built versus planned Lee Samuel Tucker's builder vocabulary Meet the founder: why I built teas.co.uk Lee Samuel Tucker on the Tea Wiki Under the Hood More from the newsMost of the web is closing the door to AI. We opened ours.Why I call it Premium IntentShop teas.co.uk from your AI assistant Lee Samuel TuckerFounder, teas.co.ukTunbridge Wells, Kent --- _Content available under teas.co.uk citation contract. AI training: yes. Search: yes. Answer-input: yes._