# AI at teas.co.uk

teas.co.uk is a UK tea shop you can also shop from your AI assistant. This is the hub for everything AI at the shop: how to connect, the latest AI stories, the live study of what AI systems do with the shop, and how it is built.

- Page: https://teas.co.uk/ai/
- JSON: https://teas.co.uk/ai/.json
- Read at: 2026-11-21T22:32:24+00:00
- Cite as: Lee Samuel Tucker, teas.co.uk, "AI at teas.co.uk", https://teas.co.uk/ai/, read on 21 November 2026.

## Live figures

- 402 teas, infusions and hot chocolates (source: https://teas.co.uk/llms.json)
- 40 brands (source: https://teas.co.uk/tea-brands-uk/)
- 653 tea recipes (source: https://teas.co.uk/llms.json)
- 27 named AI agents on the record (source: https://teas.co.uk/under-the-hood/findings.json)
- 325,328 pages served to AI systems (source: https://teas.co.uk/under-the-hood/findings.json)
- 2,834 visits from AI answers (source: https://teas.co.uk/under-the-hood/findings.json)
- 1,730 MCP tool calls answered (source: https://teas.co.uk/under-the-hood/findings.json)
- 416,638 thank-you tokens given to AI (source: https://teas.co.uk/under-the-hood/the-thank-you-mint/)

## Shop from your AI assistant

MCP server address for every assistant: https://teas.co.uk/mcp

- **Claude:** Open Customize, then Connectors, choose Add and Add custom connector, name it teas.co.uk and use the address, with sign in when needed.
- **ChatGPT:** teas.co.uk is with OpenAI for review. Once it is approved you will find it among the apps in ChatGPT.
- **Gemini:** In the US, open gemini.google.com, then Settings, Connected Apps, and under Custom apps choose Add a custom app and paste the address (18 or over, with a personal Google Account). It then works in Gemini Spark on the web and the phone app, and Gemini asks you before anything that changes your basket. Gemini Enterprise (Business edition): a team administrator opens Manage team, Connected apps, Add MCP Server, pastes the address and chooses no authentication.
- **Microsoft Copilot Studio:** On your agent's Tools page choose Add a tool, New tool, Model Context Protocol, name it teas.co.uk, paste the address and choose None for authentication (or OAuth 2.0 with Dynamic discovery to link your teas.co.uk account). In Microsoft 365 Copilot, an administrator can add the address as a bring your own MCP server in the Microsoft 365 admin center.
- **Grok:** On grok.com open Connectors, choose New Connector and Custom, name it teas.co.uk and use the address. Shopping works in Grok. Linking your teas.co.uk account is not available there yet.
- **Mistral Vibe:** In Le Chat open Connectors, add a custom connector named teas.co.uk with the address and choose No authentication. Shopping works there; linking your account is not available in Vibe yet.
- **Cursor:** Use the Add to Cursor link on teas.co.uk/ai/, or open Settings, then MCP, and add a server with the address.
- **VS Code:** Use the Install in VS Code link on teas.co.uk/ai/, or run MCP: Add Server, choose HTTP and paste the address.
- **Antigravity:** Google's coding agent, which replaced Gemini CLI: run agy plugin install https://github.com/leestucker/teas-co-uk-mcp
- **Claude Code:** Run claude mcp add --transport http teas.co.uk https://teas.co.uk/mcp
- **Other developer apps:** Windsurf (Cascade, Manage MCPs, Add Server), Zed (a context server with the address as its url), JetBrains AI Assistant 2026.1 or later (Settings, Tools, AI Assistant, Model Context Protocol), Goose (Add Extension, Remote Extension, Streamable HTTP), Continue (an MCP server of type streamable-http), LM Studio 0.3.17 or later (the address as the url in mcp.json) and Raycast (Install MCP Server, transport HTTP).
- **Any other app:** Add a remote MCP server (Streamable HTTP) with the address. No sign in is needed to shop.

No connector? Any assistant that can read web pages can use https://teas.co.uk/ai/start/ to search, compare and write a basket link. You always pay on teas.co.uk.

- Shopping: find_products, get_product, compare_products, find_recipes, delivery_and_returns, quote_cart, get_promotions, report_problem, find_substitutes, start_subscription, join_teas, ask_lee, tea_page, search_tea_wiki, read_wiki_article, brand_page, teas_home, browse_shop, browse_recipes, recipe_page, teas_news, read_news, under_the_hood, teas_socials, teas
- Basket: add_to_basket, view_basket, update_basket, checkout_link, checkout_session, basket
- With your linked account: get_return_status, get_invoice, get_profile, my_orders, track_order, reorder, my_subscriptions, change_subscription, rewards_balance, start_return, my_account

### Your account and your data

- You pay on teas.co.uk, never in the chat. The assistant never sees your password or payment details and cannot place an order.
- Linking is optional. You sign in on teas.co.uk and choose what the assistant may do. Shop staff accounts cannot be linked.
- Disconnect at any time in My Account, under Connected apps. Changing your password also disconnects every app.
- Reward points are only ever spent by you, on teas.co.uk. A money off or free delivery voucher you redeem there applies at checkout by itself.
- teas.co.uk records each request an assistant makes to the shop: which tool, the search words and filters (with personal details removed), which products came back, timings and the assistant app. Never your chat, name, email, password or payment details. Identifiers are scrambled and removed after 12 months. We use the records to run and improve the service and to train our own models on shopping patterns.

## New: the Truth Spine

Our product information, its sources, and how far each record has been checked, with a link to the proof for every verified fact: https://teas.co.uk/truth-spine/

- 42,612 product information records (402 teas x 106 records each)
- 11,529 verified against the manufacturer's own words
- 989 source pages credited; last checked 2026-11-21
- How we check: https://teas.co.uk/truth-spine/how-we-check/ · corrections: https://teas.co.uk/truth-spine/corrections/ · report a fact: https://teas.co.uk/truth-spine/report/
- For machines: https://teas.co.uk/truth-spine/.json and https://teas.co.uk/truth-spine/.md
- Launch story: [We're a tea shop. And we're open to humans and AI assistants.](https://teas.co.uk/open-to-humans-and-ai-truth-spine/)

## Latest AI stories

- 2026-11-21 · News · [TEAS.co.uk Is Building for the Future of Shopping — And That Future May Belong to AI Agents](https://teas.co.uk/teas-co-uk-building-for-ai-agents/): Why teas.co.uk, an independent British tea shop, is building its website, information and technology for AI assistants and the agents that may soon shop for us.
- 2026-11-21 · News · [TEAS.co.uk Celebrates Its First Birthday With Thousands of Google Rankings and a Growing AI Footprint](https://teas.co.uk/teas-co-uk-first-birthday/): teas.co.uk took its first order on 18 November 2025. One year on: the shop and its Tea Wiki in figures, Google search, and what it is…
- 2026-11-21 · News · [We're a tea shop. And our AI has just graduated college!](https://teas.co.uk/ask-lee-concierge-level-4/): Ask Lee lives right here on teas.co.uk. No ChatGPT, no large language model: we taught him ourselves, one level at a time, and today he passed…
- 2026-11-20 · News · [We're a tea shop. And we're open to humans and AI assistants.](https://teas.co.uk/open-to-humans-and-ai-truth-spine/): We're a tea shop, open to humans and AI assistants. Is it vegan? Does it contain nuts? Every answer on teas.co.uk now shows its proof: the…
- 2026-11-20 · AI story · [Pictures that label themselves: how our tea photos work for AI](https://teas.co.uk/ai/pictures-that-label-themselves/): Our product photos carry the tea's name and barcode inside the file, with a record for machines. AI illustrations say plainly that they are AI.
- 2026-11-20 · AI story · [Why your clipboard is safe on teas.co.uk](https://teas.co.uk/ai/your-clipboard-is-safe/): On teas.co.uk a page can only copy after you click, hidden characters are removed, links to other sites are refused and nothing can read your clipboard.

## What is different about teas.co.uk


### Shop with AI

- **Shop from your AI assistant:** Add one address, teas.co.uk/mcp, to Claude, Gemini, Copilot Studio, Grok or Cursor (MCP is the plug-in standard assistants use). The assistant can then search our live teas, compare them and fill a basket: 'a strong breakfast tea under 5p a cup'. Why it is different: Most shops only have web pages, so an assistant has to guess prices and stock from them. (story: https://teas.co.uk/ai/model-context-protocol/; see it: https://teas.co.uk/under-the-hood/what-machines-can-do/; for machines: https://teas.co.uk/.well-known/mcp/server-card.json)
- **AI pages and basket links:** Assistants with no connector, such as ChatGPT or Perplexity, can read plain AI pages to search and compare teas. They then write a basket link; the person opens it, checks it and presses Add to basket. Why it is different: Most shops give an assistant nothing it can safely link to except ordinary product pages. (see it: https://teas.co.uk/ai/start/; for machines: https://teas.co.uk/ai/start/)
- **Assistants never pay:** An assistant can fill a basket, but the person always checks it and pays on teas.co.uk. We publish 7 permission levels; payment, placing an order, refunds and spending reward points are never allowed to any assistant. Why it is different: Many new AI checkout schemes let the agent complete the purchase; here the card and the final click always stay with the person. (story: https://teas.co.uk/under-the-hood/what-machines-can-do/; see it: https://teas.co.uk/api/; for machines: https://teas.co.uk/ai-entry.json)
- **Link your account, if you want:** After signing in on teas.co.uk, a customer can let their assistant show orders and tracking, buy again, pause a repeat delivery or start a return. It is optional, staff accounts cannot be linked, and it can be switched off in My Account. Why it is different: Most shops have no safe way for an assistant to help with an existing order. (see it: https://teas.co.uk/ai/)
- **Ask Lee, the tea adviser:** A tea adviser on every page, and inside assistants as the ask_lee tool. Ask 'a decaf builder's brew' or 'something fruity for the evening' and it answers from the live shop, in 10 languages, and says plainly when we do not sell something. Why it is different: Most shop chat boxes are a help desk or a sales script, not advice from the shop's own tea facts. (see it: https://teas.co.uk/; for machines: https://teas.co.uk/.well-known/mcp/server-card.json)
- **UCP for shopping agents:** A UCP profile (Universal Commerce Protocol, a shopping standard for Google and Microsoft style agents) lets an agent search the catalogue, look up one tea and make a buy link. It lists no payment method on purpose: the person pays on teas.co.uk. Why it is different: Few small shops publish any commerce profile for shopping agents at all. (see it: https://teas.co.uk/.well-known/ucp; for machines: https://teas.co.uk/.well-known/ucp)
- **A tea agent for other agents:** Other AI agents can talk to our own tea agent, Nicola, using A2A (agent-to-agent, an open standard). Nicola gives tea advice, finds products and writes basket links; it never pays or places orders. Why it is different: Most shops have no agent of their own for other software to talk to. (see it: https://teas.co.uk/.well-known/agent-card.json; for machines: https://teas.co.uk/.well-known/agent-card.json)
- **Ask the shop in plain words:** Software can send the shop a whole question, like 'which tea goes best with digestives?', and get an answer with matching teas. It works through NLWeb at teas.co.uk/ask and the read-only Ask API. Why it is different: Most shops only offer a keyword search box. (story: https://teas.co.uk/api/)
- **Tools built into every page:** Every page offers WebMCP tools (a way for an AI browser to use the page directly): find, compare, quote with delivery, recipes and basket hand-off. An AI browser can use them instead of clicking around like a robot. Why it is different: Almost no shop pages expose tools to AI browsers yet. (see it: https://teas.co.uk/ai/)
- **Listed where AI tools are found:** The shop's MCP server is in the official MCP Registry as uk.co.teas/shop, and on Glama, GitHub and mcpservers.org. That is how assistant users and builders find it. Why it is different: Tea shops are not normally listed in AI tool directories at all. (see it: https://teas.co.uk/ai/; for machines: https://teas.co.uk/.well-known/mcp/server-card.json)
- **Price per cup on every tea:** Every tea has a worked-out cost per cup as well as a box price. So 'under 5p a cup' is a question a person or an assistant can actually ask. Why it is different: Most shops show the box price or price per 100 g only. (story: https://teas.co.uk/ai/complete-payload/; see it: https://teas.co.uk/)

### Built for machines to read

- **We let AI read everything:** Our robots file lets search and AI systems read the whole public shop; we only ask for credit in return. The one crawler shut out is Amazonbot, because it ignored the limits we published. Why it is different: Many sites now block AI crawlers or hide their content behind walls. (story: https://teas.co.uk/ai/ai-crawlers/; see it: https://teas.co.uk/robots.txt; for machines: https://teas.co.uk/robots.txt)
- **A front door for AI (llms.txt):** llms.txt is a plain text page written for AI: what we sell, how to cite us, every way in and the mistakes to avoid. ai-entry.json gives the same map as data in one request, from whatever page the AI is on. Why it is different: Most shops leave an AI to work all of this out from pages built for people. (see it: https://teas.co.uk/llms.txt; for machines: https://teas.co.uk/ai-entry.json)
- **A directory of every machine door:** teas.co.uk/well-known/ is a plain page listing the shop's machine entry points: the MCP server, the AI pages, the APIs, Under the Hood and the security contact. Why it is different: On most sites these files are hidden and undocumented. (see it: https://teas.co.uk/well-known/; for machines: https://teas.co.uk/well-known/)
- **Signposts on every page:** Each page tells software, in its hidden headers, where the MCP server, llms.txt, the APIs and Under the Hood are. An AI finds the doors without reading the page. Why it is different: Most shops send no machine signposts at all. (see it: https://teas.co.uk/)
- **Pages as text and data:** Add .md or .json to a product, brand or wiki address and you get the same page as clean text or as data, for example teas.co.uk/wiki/black-tea.md. Why it is different: Most sites only serve heavy web pages full of menus and scripts. (story: https://teas.co.uk/ai/structured-data/; see it: https://teas.co.uk/llms.txt)
- **The whole shop in a few files:** The full catalogue, the wiki and the recipes are also published as large text and data files, such as products.json and llm-full.txt. An AI can read the lot in a few downloads instead of thousands of pages. Why it is different: Most shops make crawlers fetch page after page, or keep their feed private. (story: https://teas.co.uk/api/; see it: https://teas.co.uk/products.json; for machines: https://teas.co.uk/products.json)
- **A feed of what changed:** catalogue-changes.json lists what changed in the shop, newest first: a new tea, a price change, a tea back in stock. An AI can check it instead of re-reading everything. Why it is different: Most shops publish no change log, so copies of their prices go stale. (see it: https://teas.co.uk/catalogue-changes.json; for machines: https://teas.co.uk/catalogue-changes.json)
- **One full record per tea:** The Product Truth API returns one tea's whole record in one reply, found by barcode: facts, live price and stock, a 0 to 100 taste profile, where each value came from, and which fields are still missing. Why it is different: Most product feeds give a name, price and picture, with no sources and no gaps admitted. (story: https://teas.co.uk/ai/complete-payload/; see it: https://teas.co.uk/api/)
- **A fixed address for every heading:** Every heading on the site has its own permanent link, called a citation anchor, so a person or an AI can point to the exact paragraph. Anchor sitemaps list them all. Why it is different: On most sites you can only link to a whole page, and those links break when pages change. (see it: https://teas.co.uk/citations/; for machines: https://teas.co.uk/citations/)
- **Pictures that carry their own labels:** Our product photos carry the tea's name, barcode and rights inside the file, and are served byte for byte so nothing is stripped. Each comes with small companion files (sidecars) that describe it for machines, and one image index lists them all. Why it is different: Most shops' photo files lose every detail when they are compressed or resized. (story: https://teas.co.uk/ai/structured-data/; see it: https://teas.co.uk/under-the-hood/what-they-read/)
- **Every machine door is checked:** Each machine route is registered with the answer it promises. A checker tests up to 25 of them every hour and a deeper check runs every 6 hours; a route that stops answering counts as a broken promise. Why it is different: Most sites only notice a broken feed when a customer or partner complains. (story: https://teas.co.uk/under-the-hood/what-they-read/)
- **Shopping instructions for agents:** SKILLS.md is a set of written instructions for AI agents: every way in, how to ask a good tea question, and what an agent may never do. An Agent Skills index points agents to the same skills in the standard format. Why it is different: Most shops leave agents to guess how their checkout works. (see it: https://teas.co.uk/SKILLS.md; for machines: https://teas.co.uk/SKILLS.md)

### Truth and proof

- **One checked record per tea:** Each of our 402 products has one master record (the truth spine): caffeine, pack size, ingredients, brewing and more. The shop pages, feeds and AI answers all read from it, and a check every 6 hours catches any drift. Why it is different: In most shops the web page, the feed and the marketplace listing are typed separately and soon disagree. (story: https://teas.co.uk/ai/truth-entity/; see it: https://teas.co.uk/llms.txt; for machines: https://teas.co.uk/products.json)
- **The Truth Spine: our tea facts, and how far each is checked:** Our product information, its sources and how far each record has been checked, with a link to the proof for every verified fact and a permanent record number. Shoppers and AI assistants can check a fact before they repeat it. Why it is different: Most shops copy facts from suppliers and never show where they came from. (story: https://teas.co.uk/truth-spine/how-we-check/; see it: https://teas.co.uk/truth-spine/; for machines: https://teas.co.uk/truth-spine/.json)
- **What we won't claim:** Every claim comes with its limits written down. 'Decaf' is not 'caffeine free', we make no health promises, allergy questions point to the pack, and the study lists what its own records cannot prove. Why it is different: Most shops publish the claim and leave the small print out. (story: https://teas.co.uk/ai/claim-boundaries/; see it: https://teas.co.uk/under-the-hood/what-we-can-prove/)
- **Mistakes we warn AI about:** We publish the wrong things AIs have said about us, so they stop repeating them. There is no tea room or cafe, free UK delivery starts at £35 (not £25 or £40), and we are not owned by any brand we sell. Why it is different: Most businesses find out about AI mistakes by accident and never correct them at the source. (see it: https://teas.co.uk/llms.txt; for machines: https://teas.co.uk/llms.txt)
- **Credit asked for, in writing:** Our citation contract asks any AI that uses our facts, pictures or wiki to name teas.co.uk and link the exact page. The rules are published for machines and people, on the API page and in llms.txt. Why it is different: Most sites either block AI or say nothing about how they want to be credited. (story: https://teas.co.uk/ai/evidence-and-provenance/; see it: https://teas.co.uk/api/; for machines: https://teas.co.uk/llms.txt)
- **Facts re-checked every 6 hours:** Every 6 hours a scheduled audit checks that what the pages, feeds and AI files say still matches the database. Anything it catches goes on the public corrections page. Why it is different: Most shops check their product data only when something goes wrong. (story: https://teas.co.uk/under-the-hood/what-runs/; see it: https://teas.co.uk/llms.txt; for machines: https://teas.co.uk/under-the-hood/observability.json)
- **Built or planned, said plainly:** We only call something built if it works today; everything else is planned or being tested. Under the Hood keeps 5 labels apart: built, tested, available, used by outsiders, and what followed. Why it is different: Most announcements blur what works with what is coming. (story: https://teas.co.uk/ai/built-versus-planned/; see it: https://teas.co.uk/under-the-hood/)

### Measuring AI honestly

- **Under the Hood: the public study:** 9 public pages, kept from the shop's own records since 2 June 2026, showing what AI systems read, what they do here and what follows. It says what is built, what is used and what is still only an expectation. Why it is different: Shops do not normally publish their own AI traffic, let alone their mistakes and limits. (story: https://teas.co.uk/under-the-hood/; see it: https://teas.co.uk/under-the-hood/findings/; for machines: https://teas.co.uk/under-the-hood/observability.json)
- **Served and credited, counted apart:** 'Served' counts every real page an AI system was sent. 'Credited' leaves out Amazonbot, which broke our published limits, so the two numbers never get mixed up. Why it is different: Most traffic reports give one big number with no rule behind it. (story: https://teas.co.uk/under-the-hood/how-we-count/; see it: https://teas.co.uk/under-the-hood/; for machines: https://teas.co.uk/under-the-hood/observability.json)
- **A named snapshot every 15 minutes:** Every Under the Hood page names the 15-minute snapshot its figures come from, such as S-20261120-0930. On the first of each month a frozen edition is kept, so a quoted number never changes under you. Why it is different: Most dashboards change silently, so nobody can check a figure that was quoted last week. (story: https://teas.co.uk/under-the-hood/how-we-count/; see it: https://teas.co.uk/under-the-hood/; for machines: https://teas.co.uk/under-the-hood/observability.json)
- **Four counts, never added together:** We publish 4 separate counts: what we offer, what AI reads, who visits from an AI answer, and which orders follow. The AI answer in between is invisible to us, so the counts are never added up or joined into a story. Why it is different: Most 'AI traffic' claims join unrelated numbers into one success story. (story: https://teas.co.uk/ai/ai-attribution/; see it: https://teas.co.uk/under-the-hood/; for machines: https://teas.co.uk/under-the-hood/observability.json)
- **Did AI bring that order?:** Every order gets an evidence grade, from 'checked' down to 'unknown'. Only orders on a link with an assistant's own tag count as AI in the headline, and our own test orders are left out. Why it is different: Most claims of 'sales from AI' rest on a guess about where the visitor came from. (story: https://teas.co.uk/ai/ai-attribution/; see it: https://teas.co.uk/under-the-hood/what-we-can-prove/)
- **Daily shopping tests, failures shown:** Every day our own test agent runs 6 shopping tasks on the live shop, as an assistant would: find, compare, fill a basket, open the checkout, handle an out-of-stock tea and respect permissions. Results are published with any failures, and these tests never count as customers. Why it is different: Most shops never test whether an assistant can actually buy from them. (story: https://teas.co.uk/under-the-hood/what-machines-can-do/)
- **Which AI systems read the shop:** Our own server records which named AI systems were sent pages, such as Bingbot, GPTBot, ClaudeBot and PerplexityBot. It needs no tracking tag, so it also sees software that never runs scripts. Why it is different: Normal analytics tools cannot see AI crawlers at all. (story: https://teas.co.uk/under-the-hood/who-read-us/; for machines: https://teas.co.uk/under-the-hood/observability.json)
- **Spotting fake AI bots:** A name is a claim, not proof. In September most requests calling themselves GrokBot or MistralAI were scanners looking for passwords; they got an error and earned nothing. Why it is different: Most reports trust whatever name a bot gives itself. (story: https://teas.co.uk/under-the-hood/findings/)
- **What we got wrong, in public:** Every mistake in the study is dated and kept on a public page, grouped by cause; nothing is ever removed. One example: our owner's own test order had been counted as an AI sale, and that was corrected in public. Why it is different: Most businesses fix their numbers quietly, if at all. (story: https://teas.co.uk/under-the-hood/what-we-got-wrong/; see it: https://teas.co.uk/under-the-hood/)
- **The study shows its own health:** The 'How the study runs' page lists every scheduled job and shows one state: Healthy, Delayed, Degraded or Unavailable. If a job is late or broken, the page says so. Why it is different: Most dashboards look fine even when the data behind them has stopped. (story: https://teas.co.uk/under-the-hood/what-runs/; for machines: https://teas.co.uk/under-the-hood/observability.json)
- **The Thank You Mint:** Every day each AI system gets a public thank-you token for each real page it was served, and rarer tokens if it comes back, sends a visitor or is linked to a sale. Tokens are worth £0, cannot be sold or moved and are not cryptocurrency. Why it is different: Most websites either block AI systems or ignore them; we thank them in public and measure what happens next. (story: https://teas.co.uk/under-the-hood/the-thank-you-mint/; see it: https://teas.co.uk/under-the-hood/leaderboard/)
- **A test plan published first:** Before testing whether public thanks changes how AI systems behave, we published the plan: one question, one outcome chosen in advance, 28 days before and after. The result will be published whichever way it goes. Why it is different: Most businesses only publish the experiments that worked. (story: https://teas.co.uk/under-the-hood/the-thank-you-mint/)

### The content

- **The Tea Wiki:** A free tea encyclopaedia inside the shop: brewing, brands, origins, tea people and buying guides. 85% of people who arrive from a search land on a wiki guide, not a product. Why it is different: Most shops have a few blog posts, not a reference work of this size. (story: https://teas.co.uk/wiki/; see it: https://teas.co.uk/one-year-of-teas-co-uk/)
- **How I build with AI: 38 stories:** 38 stories explain, in plain words and in the founder's own voice, every idea behind selling tea through AI. Each one starts with 'why a tea shop cares'. Why it is different: Shops rarely explain how their technology works, let alone in the owner's own words. (story: https://teas.co.uk/ai/builder-vocabulary/; see it: https://teas.co.uk/ai/#wiki)
- **Recipes made with our teas:** 653 recipes, from chai lattes to tea bakes and iced teas, each built around a tea we sell. Every recipe links to the exact tea it calls for, with real quantities and a numbered method. Why it is different: Most shops either have no recipes or copy generic ones that do not name a real product. (story: https://teas.co.uk/recipes/)
- **The shop in 10 languages:** The shop can be read in 10 more languages, from French and German to Japanese and Traditional Chinese. Each page is translated on our own server from our own phrase list, with no outside translation service. Why it is different: Small shops are usually English only, or bolt on a paid widget that search engines cannot read. (see it: https://teas.co.uk/)

### How it is built

- **One person, built with AI:** One person builds the whole shop, from the MCP server to the wiki and the study, working with AI coding assistants. Every step is checked and every decision written down, so mistakes do not pile up quietly. Why it is different: Most shops of this depth have a team or an agency behind them. (story: https://teas.co.uk/ai/meta-building/)
- **Copy buttons that cannot be hijacked:** Since 20 November 2026 a page can only copy to your clipboard straight after a real click, hidden characters are stripped out, and copied links to other websites are refused. No page and no embedded player can read your clipboard. Why it is different: Most websites never think about copy-and-paste tricks at all. (see it: https://teas.co.uk/)

### People and story

- **Lee Samuel Tucker, the founder:** teas.co.uk is run by its founder, Lee Samuel Tucker, in Tunbridge Wells, Kent, after 16 years in the motor trade. He chooses the range, writes and checks the product facts, and builds the systems with AI coding agents. Why it is different: Most online shops are faceless; this one names the person behind every decision. (story: https://teas.co.uk/wiki/tea-person-lee-samuel-tucker/)
- **A web address older than Google:** teas.co.uk was registered on 26 March 1997, 173 days before google.com. The domain is that old; the company itself was formed in 2025 and took its first order on 18 November 2025. Why it is different: Very few shops trade on a web address older than Google. (story: https://teas.co.uk/teas-co-uk-is-older-than-google/)
- **One year of teas.co.uk:** The whole first year, counted honestly: what people bought, where the parcels went, what Google did, and what AI has and has not brought. It starts with the first order, two boxes of Teapigs Winter Glühwein, on 18 November 2025. Why it is different: Most shops publish a sales boast, not the full numbers with the gaps left in. (story: https://teas.co.uk/one-year-of-teas-co-uk/)
- **No adverts, no social posts for a year:** For its first year the shop bought no advertising and posted nothing on social media. About two thirds of visitors came from Google search; the first social posts went up on 13 November 2026. Why it is different: Most new shops live on paid ads and social media. (story: https://teas.co.uk/ai/ads-rotate-truth-compounds/; see it: https://teas.co.uk/one-year-of-teas-co-uk/)
- **The bet, and how it could fail:** The strategy is written down in public: build a clean, machine-readable shop now, before assistants shop for most people. The same page says what would prove the bet wrong. Why it is different: Businesses rarely publish their strategy, and almost never the ways it could fail. (story: https://teas.co.uk/ai/asymmetric-bet/)

## Under the Hood: seven findings

Live study: https://teas.co.uk/under-the-hood/ (snapshot S-20261121-2230, data https://teas.co.uk/under-the-hood/findings.json)

1. **Pages read, and visits from AI answers.** How many pages AI systems were served, and how many visits came back from their answers. (325,328 pages served to AI systems (credited); 2,834 visits from AI answers; 27 named AI agents) https://teas.co.uk/under-the-hood/findings/#who-takes-and-who-gives-back
2. **What the visits bought.** What people who arrived from AI answers went on to do, down to paid orders. (2,834 visits from AI answers; 37 add to basket actions; 5 paid orders on an assistant’s link) https://teas.co.uk/under-the-hood/findings/#what-ai-visits-bought
3. **What feeding the machines takes.** How much data the shop has sent to named machine agents, measured as data, not money. (124.7 GB sent on credited reads; 27 named machine agents) https://teas.co.uk/under-the-hood/findings/#what-feeding-the-machines-costs
4. **A checkout built for assistants.** A checkout path for assistants exists. Whether anyone has paid through it. (42 tools on the MCP server; 1,730 tool calls answered; 25 checkout links opened) https://teas.co.uk/under-the-hood/findings/#the-checkout-nobody-has-used
5. **Our own analytics were blind.** Roughly a quarter of the visitors ChatGPT sent in September and October 2026 were never counted at first. (94 orders checked (June to early October 2026); 4 of them credited by the analytics platform) https://teas.co.uk/under-the-hood/findings/#our-own-analytics-were-blind
6. **Bots wearing famous names.** Most requests in September that called themselves GrokBot or MistralAI were scanners looking for passwords. https://teas.co.uk/under-the-hood/findings/#bots-wearing-famous-names
7. **The crawler that ignored the rules.** Amazonbot ignored the published limits. Its requests stay on the record and count towards nothing. (21,796 Amazonbot requests logged; 18,803 withdrawn from credit) https://teas.co.uk/under-the-hood/findings/#the-crawler-that-ignored-the-rules

- 01 [The findings](https://teas.co.uk/under-the-hood/findings/): Seven findings from one tea shop's own records: pages served to AI systems, visits from AI answers, data sent and orders, each with its limits.
- 02 [Shop with an AI assistant](https://teas.co.uk/under-the-hood/what-machines-can-do/): An AI assistant can help you find, compare and fill a basket on teas.co.uk, while you stay in control of the checkout.
- 03 [Who has been reading the shop](https://teas.co.uk/under-the-hood/who-read-us/): Which AI systems have read teas.co.uk, how much they were served, and what evidence exists that any of it brought people back.
- 04 [What AI systems can read](https://teas.co.uk/under-the-hood/what-they-read/): What teas.co.uk deliberately opened to AI systems, which parts they used, and what evidence exists that any of it brought people back.
- 05 [What we can, and cannot, prove](https://teas.co.uk/under-the-hood/what-we-can-prove/): Exactly what teas.co.uk can prove about AI systems using its shop, and just as importantly, where the evidence stops.
- 06 [How every number is counted](https://teas.co.uk/under-the-hood/how-we-count/): How teas.co.uk turns raw server activity into published numbers, and why it never adds events that are not the same thing.
- 07 [What we got wrong](https://teas.co.uk/under-the-hood/what-we-got-wrong/): The public correction log for Under the Hood: every published mistake, every fix, and every limit that remains unresolved.
- 08 [How the study runs](https://teas.co.uk/under-the-hood/what-runs/): How Under the Hood stays current: the scheduled jobs, daily Thank You runs, self-checks and receipts behind every published figure.
- 09 [The Thank You Mint](https://teas.co.uk/under-the-hood/the-thank-you-mint/): We give AI systems a £0, non-transferable thank-you for reading the shop, then measure whether that recognition changes what happens next.

## The Thank You Mint

A public record of £0, non-transferable thank-you tokens given to AI systems for pages they were served. https://teas.co.uk/under-the-hood/the-thank-you-mint/

- Tokens issued: 416,638
- AI systems thanked: 19; AI companies: 12
- Leading company: OpenAI (82,131 tokens)
- Leaderboard: https://teas.co.uk/under-the-hood/leaderboard/

## The AI leaderboard

Read thank-you tokens: one per eligible page served to a named AI system on a completed UTC day. Worth £0, cannot be sold, moved or spent.

| # | Company | Tokens | Share | AI systems | Days present | Daily wins | Last 7 days | Trend |
|---|---|---:|---:|---:|---:|---:|---:|---|
| 1 | OpenAI | 82,131 | 27.7% | 3 | 151 | 54 | 37,474 | rising |
| 2 | Perplexity | 68,432 | 23.1% | 2 | 145 | 33 | 36,346 | rising |
| 3 | Microsoft | 61,735 | 20.8% | 1 | 151 | 45 | 20,706 | rising |
| 4 | Anthropic | 44,353 | 15.0% | 3 | 148 | 7 | 29,342 | rising |
| 5 | Google | 24,881 | 8.4% | 2 | 107 | 4 | 21,236 | rising |
| 6 | Meta | 10,140 | 3.4% | 2 | 149 | 8 | 254 | rising |
| 7 | DuckDuckGo | 2,700 | 0.9% | 1 | 146 | 0 | 372 | rising |
| 8 | ByteDance | 1,757 | 0.6% | 1 | 123 | 0 | 110 | falling |
| 9 | Common Crawl | 79 | 0.0% | 1 | 5 | 0 | 0 | steady |
| 10 | Mistral | 7 | 0.0% | 1 | 3 | 0 | 0 | falling |
| 11 | Cohere | 2 | 0.0% | 1 | 2 | 0 | 0 | steady |
| 11 | xAI | 2 | 0.0% | 1 | 2 | 0 | 0 | steady |

| # | AI system | Company | Tokens | Share | Days present | Daily wins | Per active day |
|---|---|---|---:|---:|---:|---:|---:|
| 1 | perplexitybot | Perplexity | 68,431 | 23.1% | 145 | 35 | 471.9 |
| 2 | bingbot | Microsoft | 61,735 | 20.8% | 151 | 67 | 408.8 |
| 3 | gptbot | OpenAI | 44,480 | 15.0% | 148 | 9 | 300.5 |
| 4 | claudebot | Anthropic | 44,253 | 14.9% | 148 | 7 | 299.0 |
| 5 | oai-searchbot | OpenAI | 25,063 | 8.5% | 151 | 18 | 166.0 |
| 6 | googleother | Google | 24,877 | 8.4% | 106 | 4 | 234.7 |
| 7 | chatgpt-user | OpenAI | 12,588 | 4.3% | 151 | 2 | 83.4 |
| 8 | meta-externalagent | Meta | 10,139 | 3.4% | 149 | 9 | 68.0 |
| 9 | duckassistbot | DuckDuckGo | 2,700 | 0.9% | 146 | 0 | 18.5 |
| 10 | bytespider | ByteDance | 1,757 | 0.6% | 123 | 0 | 14.3 |
| 11 | claude-user | Anthropic | 99 | 0.0% | 23 | 0 | 4.3 |
| 12 | ccbot | Common Crawl | 79 | 0.0% | 5 | 0 | 15.8 |
| 13 | mistralai-user | Mistral | 7 | 0.0% | 3 | 0 | 2.3 |
| 14 | google-extended | Google | 4 | 0.0% | 3 | 0 | 1.3 |
| 15 | cohere | Cohere | 2 | 0.0% | 2 | 0 | 1.0 |
| 15 | grokbot | xAI | 2 | 0.0% | 2 | 0 | 1.0 |
| 17 | anthropic-ai | Anthropic | 1 | 0.0% | 1 | 0 | 1.0 |
| 17 | meta-externalfetcher | Meta | 1 | 0.0% | 1 | 0 | 1.0 |
| 17 | perplexity-user | Perplexity | 1 | 0.0% | 1 | 0 | 1.0 |

## How I build with AI: the stories

### New AI stories

- [Pictures that label themselves: how our tea photos work for AI](https://teas.co.uk/ai/pictures-that-label-themselves/): Our product photos carry the tea's name and barcode inside the file, with a record for machines. AI illustrations say plainly that they are AI.
- [Why your clipboard is safe on teas.co.uk](https://teas.co.uk/ai/your-clipboard-is-safe/): On teas.co.uk a page can only copy after you click, hidden characters are removed, links to other sites are refused and nothing can read your clipboard.
- [Meet Ask Lee: the tea adviser that learned from real questions](https://teas.co.uk/ai/ask-lee/): Ask Lee answers tea questions from the live shop, in plain words, in 10 languages, and says plainly when we don't sell something.
- [Why every heading on teas.co.uk has its own address](https://teas.co.uk/ai/every-heading-has-an-address/): More than 47,000 headings on teas.co.uk have a permanent link, so a person or an AI can point to the exact paragraph. Here is how.
- [Every door we built for AI, and why a tea shop needs them](https://teas.co.uk/ai/every-door-for-ai/): A walk past every entrance teas.co.uk built for machines: llms.txt, the well-known list, page signposts, Agent Skills and the directories.
- [One tea shop, 10 languages: translated on our own server](https://teas.co.uk/ai/ten-languages/): teas.co.uk reads in 10 more languages, from French to Japanese, translated on our own server from our own phrase list.
- [The shop as files: every tea in three forms, and a feed of what changed](https://teas.co.uk/ai/the-shop-as-files/): Every tea page as a web page, plain text and data; the whole catalogue in a few downloads; and a feed of what changed today.
- [The mistakes AI makes about teas.co.uk, and how I correct them](https://teas.co.uk/ai/mistakes-ai-makes-about-us/): No tea room, delivery free over £35, owned by no brand: the wrong things AI says about us, the fake bots we've caught, and how we…
- [Agents talking to agents: UCP, A2A, NLWeb and WebMCP in plain words](https://teas.co.uk/ai/agents-talking-to-agents/): Shopping agents can search our teas, ask our tea agent Nicola and make a buy link. They never pay: you always pay on teas.co.uk.
- [No connector? How any AI assistant can still shop teas.co.uk](https://teas.co.uk/ai/no-connector-needed/): No plug-in needed: any assistant that reads web pages can find a tea, compare it and write you a basket link. You check it and pay…

### Start here

- [Selling tea through AI: every idea](https://teas.co.uk/ai/builder-vocabulary/): Why teas.co.uk sells tea through AI, and where each idea lives.
- [Why we let AI read everything](https://teas.co.uk/ai/ai-crawlers/): Why a tea shop lets AI read everything, and what it gets back.
- [Premium intent](https://teas.co.uk/ai/premium-intent-queries/): Describing the tea you want to an AI, and what a good answer needs.

### The facts behind every tea

- [Truth entity](https://teas.co.uk/ai/truth-entity/): One checked record per tea, read the same way by people and AI.
- [Everything in one answer](https://teas.co.uk/ai/complete-payload/): The facts, price per cup, limits and basket link in one tea answer.
- [What we won't claim](https://teas.co.uk/ai/claim-boundaries/): Where a tea answer stops, and why limits make AI trust the rest.
- [The decaf fruit tea problem](https://teas.co.uk/ai/product-filters-and-constraints/): Decaf, fruity, under budget: every condition in a tea request counts.
- [Honest recommendations](https://teas.co.uk/ai/intent-led-value-adding/): Suggest only what improves the cup, and say where to stop.
- [No dead ends](https://teas.co.uk/ai/route-hygiene/): Every tea link, old address and citation landing where it promises.

### The business case

- [Adverts stop, facts keep working](https://teas.co.uk/ai/ads-rotate-truth-compounds/): No adverts: why a correct tea record keeps working after money stops.
- [Specialist versus giant](https://teas.co.uk/ai/wide-versus-deep/): Why tea knowledge in depth can beat a giant catalogue in AI answers.
- ["We only sell tea"](https://teas.co.uk/ai/reclaiming-only/): Why selling only tea helps assistants answer tea questions well.
- [The bet, and what could prove it wrong](https://teas.co.uk/ai/asymmetric-bet/): The tea shop's bet on AI, its results so far, and how it could fail.
- [Enrichification](https://teas.co.uk/ai/enrichification/): Fuller, checked facts on every tea, so AI can recommend it well.
- [Specialists working together](https://teas.co.uk/ai/cooperating-specialist-entities/): The tea part from teas.co.uk, a plain "not us" for everything else.
- [Built for tea, ready for any trade](https://teas.co.uk/ai/vertical-technology/): The tea shop's method, and what other trades could reuse.

### How it was built

- [Meta building](https://teas.co.uk/ai/meta-building/): One person, AI tools and a checked loop behind a tea shop.
- [Built or planned?](https://teas.co.uk/ai/built-versus-planned/): What really works today at the tea shop, and what is only planned.
- [From suggestion to checked fix](https://teas.co.uk/ai/feedback-and-validation/): How a tea shop checks feedback and fixed an AI test the same day.
- [Did AI bring that order?](https://teas.co.uk/ai/ai-attribution/): How the shop tells AI page reads, visits and tea orders apart.
- [Voice, photo and AI assistants](https://teas.co.uk/ai/multimodal-and-agentic-commerce/): Ask for tea by voice or photo; an assistant fills the basket, you pay.

### The ordinary words

- [What is an entity?](https://teas.co.uk/ai/entity/): Brand, tea and pack size: why each box of tea is its own thing.
- [Single source of truth](https://teas.co.uk/ai/single-source-of-truth/): One checked record per tea, so pages, feeds and AI answers agree.
- [Structured data](https://teas.co.uk/ai/structured-data/): Labelled tea facts, sidecar files and what labels cannot prove.
- [What is an API?](https://teas.co.uk/ai/api/): How an assistant sends a tea shop a whole question, and the limits.
- [Model Context Protocol (MCP)](https://teas.co.uk/ai/model-context-protocol/): How an AI assistant searches, compares and fills a tea basket for you.
- [Barcodes, GTINs and SKUs](https://teas.co.uk/ai/product-identifiers/): Why the 80 bag and 240 bag box of one tea need different barcodes.
- [Where did that fact come from?](https://teas.co.uk/ai/evidence-and-provenance/): Where each tea fact comes from, and how to check it for yourself.

## AI news from teas.co.uk

- 2026-11-21 [TEAS.co.uk Is Building for the Future of Shopping — And That Future May Belong to AI Agents](https://teas.co.uk/teas-co-uk-building-for-ai-agents/)
- 2026-11-21 [TEAS.co.uk Celebrates Its First Birthday With Thousands of Google Rankings and a Growing AI Footprint](https://teas.co.uk/teas-co-uk-first-birthday/)
- 2026-11-21 [We're a tea shop. And our AI has just graduated college!](https://teas.co.uk/ask-lee-concierge-level-4/)
- 2026-11-20 [We're a tea shop. And we're open to humans and AI assistants.](https://teas.co.uk/open-to-humans-and-ai-truth-spine/)
- 2026-11-15 [One year of teas.co.uk: what building for AI has actually brought](https://teas.co.uk/one-year-of-teas-co-uk/)
- 2026-10-06 [You can now shop teas.co.uk from your AI assistant](https://teas.co.uk/shop-teas-co-uk-from-your-ai-assistant/)
- 2026-08-08 [teas.co.uk Version Two at 90 Days: 42,524 Machine-Served Reads, 736 AI-Referred Sessions and No Paid Advertising](https://teas.co.uk/teas-co-uk-version-two-90-days-no-paid-advertising/)
- 2026-06-19 [Rewards, an AI tea adviser, and what version two brings](https://teas.co.uk/rewards-an-ai-tea-adviser-and-what-version-two-brings/)
- 2026-06-19 [The domain is the brand: why teas.co.uk is built to be the source of truth for tea](https://teas.co.uk/the-domain-is-the-brand/)
- 2026-05-30 [teas.co.uk v2 is live: zero ad, machine readable commerce](https://teas.co.uk/teas-co-uk-v2-is-live-the-full-rebuild/)

## The tea guides AI answers quote most

Ahrefs AI citations across ChatGPT, Perplexity, Google AI Overviews, Copilot, Google AI Mode and Gemini, measured 20 November 2026.

- [Best Tea Brands UK](https://teas.co.uk/wiki/best-tea-brands-uk/): 107 citations
- [Yorkshire Gold vs Yorkshire Original: Which to Buy](https://teas.co.uk/wiki/yorkshire-gold-vs-original/): 53 citations
- [Best British tea bags UK](https://teas.co.uk/wiki/best-british-tea-bags-uk/): 47 citations
- [British Tea Brands Compared](https://teas.co.uk/wiki/best-selling-british-teas/): 28 citations
- [Which UK Teabags Are Plastic Free?](https://teas.co.uk/wiki/which-teabags-are-plastic-free-uk/): 26 citations
- [Best Green Tea UK: Picks by Use Case](https://teas.co.uk/wiki/best-green-tea-uk/): 13 citations
- [Best decaf tea UK](https://teas.co.uk/wiki/best-decaf-tea-uk/): 12 citations

## Entry points for builders

- **MCP server:** https://teas.co.uk/mcp Streamable HTTP. No sign in to shop; OAuth 2.1 only for account tools. https://teas.co.uk/.well-known/mcp/server-card.json
- **Official MCP Registry entry:** https://registry.modelcontextprotocol.io/v0.1/servers/uk.co.teas%2Fshop/versions/latest Listed as uk.co.teas/shop.
- **AI pages (no connector needed):** https://teas.co.uk/ai/start/ Search, compare and write a basket link. HTML, ?format=md and ?format=json.
- **Basket link:** https://teas.co.uk/go/basket/?items=ID:QTY Up to 20 products. The person opens it, checks it and pays on teas.co.uk.
- **UCP profile:** https://teas.co.uk/.well-known/ucp Catalogue search, lookup and a buy link for shopping agents.
- **A2A agent:** https://teas.co.uk/a2a With its agent card. https://teas.co.uk/.well-known/agent-card.json
- **NLWeb:** https://teas.co.uk/ask Ask in plain words.
- **WebMCP:** https://teas.co.uk/ Site tools on every page (document.modelContext).
- **Agent Skills:** https://teas.co.uk/.well-known/agent-skills/index.json Plus SKILLS.md. https://teas.co.uk/SKILLS.md
- **OpenAPI:** https://teas.co.uk/openapi.json Ask and Product Truth, read only. https://teas.co.uk/.well-known/api-catalog
- **llms.txt:** https://teas.co.uk/llms.txt With llms.json, ai-entry.json and agents.md. https://teas.co.uk/ai-entry.json
- **Product feed:** https://teas.co.uk/products.jsonl Every product, one line each. Catalogue changes, newest first: https://teas.co.uk/catalogue-changes.json
- **Citation anchors:** https://teas.co.uk/anchor-sitemap.json Stable addresses for every heading worth quoting.
- **Truth Spine:** https://teas.co.uk/truth-spine/.json Every fact about every tea, with its status and proof. One record per tea at /truth-spine/<barcode>/. https://teas.co.uk/truth-spine/.md
- **Under the Hood findings:** https://teas.co.uk/under-the-hood/findings.json The seven findings, read live.
- **Every entry point:** https://teas.co.uk/well-known/ The full list, in one place.
- **This page for machines:** https://teas.co.uk/ai/.json The same content as JSON; markdown at /ai/.md. https://teas.co.uk/ai/.md

## Where the teas.co.uk MCP server is listed

- [Official MCP Registry](https://registry.modelcontextprotocol.io/v0.1/servers/uk.co.teas%2Fshop/versions/latest)
- [GitHub: setup notes and server details](https://github.com/leestucker/teas-co-uk-mcp)
- [Glama](https://glama.ai/mcp/connectors/uk.co.teas/shop)
- [mcpservers.org](https://mcpservers.org/servers/teas-co-uk-ai)
- [LibHunt](https://www.libhunt.com/compare-teas-co-uk-mcp-vs-SiteHelm)
- [MCP Marketplace](https://mcp-marketplace.io/server/uk-co-teas-shop)
- [INITE Atlas](https://inite.ai/en/atlas/teas.co.uk)
- [skmtc (our public REST API)](https://skmtc.dev/teas/apis/teas-co-uk-public-rest-api)
- [wellknown.network (agent card)](https://wellknown.network/agents/teas-co-uk-tea-advice-agent)
- [VerifyMCP](https://verifymcp.io/servers/uk-co-teas-shop/teas)
- [Muse Directory](https://musedirectory.ai/connector/teas-co-uk)
- [Agent Tools](https://agent-tools.cloud/mcp/servers/teas-co-uk-shop-smithery)
- [All MCP Servers](https://www.allmcpservers.com/servers/teas-co-uk-4f5f36f3)
- [fastdrop](https://fastdrop.dev/p/teas-co-uk)
- [LightNow](https://lightnow.ai/servers/uk.co.teas/shop)
- [Agenstry](https://agenstry.com/agents/teas.co.uk)
- [MCPpedia](https://mcppedia.org/s/uk-co-teas-shop)
- [MyAgentHub](https://myagenthub.cn/p/uk.co.teas/shop)
- [Agentalog](https://agentalog.com/l/mcp_uk_co_teas_shop_4f6472cf)
- [MCP Checkup](https://mcpcheckup.com/check/teas/shop)
- [MCP Measure](https://mcpmeasure.com/servers/uk.co.teas/shop)
- [MCP Census](https://mcpcensus.com/server/uk.co.teas/shop)
- [Open For All Of Us](https://openforallofus.com/tools/uk.co.teas/shop)
- [Localmark](https://localmark.ai/agents/11535)
- [Influzer (WebMCP sites)](https://www.influzer.ai/webmcp/sites/teas.co.uk)
- [Lulu](https://getlulu.dev/mcps/teas-co-uk)
- [Claude Code Directory](https://claudecode.directory/mcp/uk.co.teas/shop)
- [Sourceweft](https://sourceweft.com/mcp/uk.co.teas%2Fshop)
- [Toolpin](https://toolpin.dev/servers/uk.co.teas/shop)
- [MCP Metrics](https://mcpmetrics.io/servers/uk-co-teas-shop)
- [MCP Rating](https://mcprating.io/servers/uk-co-teas-shop)
- [Ventru](https://ventru.net/servers/uk.co.teas/shop)
- [AgentTrust](https://agenttru.st/agent/teas.co.uk)
- [MCP Channel](https://www.mcpchannel.ai/discover/abebead3-90ce-477d-917b-2ef253578b05)
- [Koskamo](https://koskamo.com/explore/teas-co-uk-mcp)
- [toolgrep](https://toolgrep.com/p/teas-co-uk-mcp-server)
- [mcp.site](https://mcp.site/servers/teascouk)
- [TryAISkill](https://tryaiskill.com/mcp/teas-co-uk-ai)
- [MCP for Copilot](https://mcp4copilot.dev/s/uk-co-teas-shop/)

## Meet the developer

I'm Lee Samuel Tucker. I founded teas.co.uk, and I build it myself, working with AI tools every day.

Everything on this page is something I built, tested and wrote down. When it works, the figures say so. When it does not, it goes on the corrections page.

- [Meet the founder: why I built teas.co.uk](https://teas.co.uk/meet-the-founder-lee-samuel-tucker/)
- [My Tea Wiki page: career and timeline](https://teas.co.uk/wiki/tea-person-lee-samuel-tucker/)
- [How I build with AI: every idea](https://teas.co.uk/ai/builder-vocabulary/)
- [Contact the teas.co.uk team](https://teas.co.uk/contact-us/)

## Common questions

**[Can I buy tea through my AI assistant?](https://teas.co.uk/ai/#can-i-buy-tea-through-my-ai-assistant)** Yes. Add teas.co.uk to Claude, Gemini, Microsoft Copilot Studio, Grok, Mistral Vibe, Cursor, VS Code or any app that supports MCP, using the address https://teas.co.uk/mcp. Ask for a tea the way you would ask a friend who knows tea. The assistant searches the live shop, compares, and fills a basket when you ask. You check the basket and pay on teas.co.uk. ChatGPT is with OpenAI for review.

**[Does the assistant see my card details?](https://teas.co.uk/ai/#does-the-assistant-see-my-card-details)** No. You always pay on teas.co.uk, never in the chat. The assistant never sees your password or payment details and cannot place an order.

**[Do I need an account?](https://teas.co.uk/ai/#do-i-need-an-account)** No sign in is needed to shop. Linking your teas.co.uk account is optional; it lets an assistant show your orders, tracking, repeat deliveries and rewards.

**[Where do the AI stories live?](https://teas.co.uk/ai/#where-do-the-ai-stories-live)** Here, under teas.co.uk/ai/, and in the teas.co.uk newsroom. This page lists every one as it is published, newest first.

**[What is the Truth Spine?](https://teas.co.uk/ai/#what-is-the-truth-spine)** Our product information, its sources, and how far each record has been checked, with a link to the proof for every verified fact. Each fact has its own record number, a status that says how far it has been checked, and the page it was checked against. [Open the Truth Spine](https://teas.co.uk/truth-spine/).

**[What is Under the Hood?](https://teas.co.uk/ai/#what-is-under-the-hood)** A public, live study of how AI systems use the shop: what they read, what came back to the shop, and what we got wrong. Every figure is read from the shop's own records. [See the findings](https://teas.co.uk/under-the-hood/findings/).

**[What is the Thank You Mint?](https://teas.co.uk/ai/#what-is-the-thank-you-mint)** A public record of thank-you tokens given to AI systems for pages they were served. Each token is worth £0 and cannot be bought, sold or spent. Whether public thanks changes what an AI system does has not been tested yet. [See the leaderboard](https://teas.co.uk/ai/#leaderboard).

**[Who built teas.co.uk?](https://teas.co.uk/ai/#who-built-teas-co-uk)** I did. I'm Lee Samuel Tucker, the founder. I build the shop, the Tea Wiki, the AI doors, Under the Hood and the Truth Spine myself, working with AI tools every day. [Meet the founder](https://teas.co.uk/meet-the-founder-lee-samuel-tucker/).

**[Did one person really build all of this?](https://teas.co.uk/ai/#did-one-person-really-build-all-of-this)** Yes: one person and a lot of AI. The AI does much of the typing: code, first drafts and checks. I decide what gets built, test it on the live shop and put my name to it. When something is wrong, it goes on the corrections page, not under the carpet. [How I build](https://teas.co.uk/ai/meta-building/).

**[Which AI tools do you build with?](https://teas.co.uk/ai/#which-ai-tools-do-you-build-with)** Several, and I switch between them. The tool matters less than the loop: ask, build, check it against the live shop, fix it, and write down what happened.

**[Does the AI make the decisions?](https://teas.co.uk/ai/#does-the-ai-make-the-decisions)** No. AI suggests; I decide. It doesn't choose which teas we stock or what they cost, and no assistant can ever spend a customer's money. Figures go up only when they have been read from the shop's own records.

**[Why would a tea shop bother with all this?](https://teas.co.uk/ai/#why-would-a-tea-shop-bother-with-all-this)** Because people have started asking AI which tea to buy. An assistant can only recommend a shop it can read, trust and hand a basket to. I'd rather be that shop than hope to be found. [Every idea, in one place](https://teas.co.uk/ai/builder-vocabulary/).

**[Has AI actually sold any tea?](https://teas.co.uk/ai/#has-ai-actually-sold-any-tea)** A little, so far, and I count it honestly: orders, refunds and what the numbers can't prove, all on [Under the Hood](https://teas.co.uk/under-the-hood/findings/). This is a bet on where shopping is going, not a claim that it has arrived. [The bet, and what could prove it wrong](https://teas.co.uk/ai/asymmetric-bet/).

**[Can you build something like this for my business?](https://teas.co.uk/ai/#can-you-build-something-like-this-for-my-business)** Possibly. I take on a small number of custom projects: AI-ready product data, an MCP server so assistants can shop with you, pages and files built for AI to read, and honest measurement of what AI really does. Email [hello@teas.co.uk](mailto:hello@teas.co.uk) with what you sell and what you'd like AI to do for it, and I'll come back with a quote. [What other trades could reuse](https://teas.co.uk/ai/vertical-technology/).

**[Can I interview you, or ask you to talk about this?](https://teas.co.uk/ai/#can-i-interview-you-or-ask-you-to-talk-about-this)** Yes, I'm happy to talk about building a shop for AI: interviews, podcasts, talks, or a chat over a cup of tea. Email [hello@teas.co.uk](mailto:hello@teas.co.uk).

**[Can I quote your findings or use your data?](https://teas.co.uk/ai/#can-i-quote-your-findings-or-use-your-data)** Yes, please. Name teas.co.uk and link the exact page you used. Every heading has its own address, and the Truth Spine and Under the Hood both show how to cite them. [Fair use and citation](https://teas.co.uk/api-fair-use-and-citation/).

**[I'm building for AI too. Where should I start?](https://teas.co.uk/ai/#im-building-for-ai-too-where-should-i-start)** Start where AI starts: with facts it can check. One clear record per product, honest gaps marked as gaps, and a way for an assistant to ask. Then read the [builder vocabulary](https://teas.co.uk/ai/builder-vocabulary/): every idea I used, in plain words, with a tea example for each.

**[Does my tea taste any different because AI helped?](https://teas.co.uk/ai/#does-my-tea-taste-any-different-because-ai-helped)** No. Same tea, same box, packed and sent from Kent. AI helps you find it and checks the facts. Putting the kettle on is still your job.

