Showing 1–12 of 14 results
Showing 1–12 of 14 results
A Busymate AI demonstration
Fernweh Supply Co. is invented and sells nothing — the WooCommerce store under it is real, and so is every capability below. Open the chat and try the line under each one.
The assistant is grounded in this store's own catalogue and its shipping, returns, privacy and terms pages — and nothing else. Ask it something only this shop knows and it answers with the real number; ask it something outside and it says so rather than inventing one.
Try it “What is the Wanderweg made of, and how much does it weigh empty?”
Six of this page's own actions are published to the assistant over WebMCP — search the range, read the cart, add a line, take one out, open a product, and check who is signed in. They run in your browser, in your session, through the same WooCommerce endpoints the buttons use, so nothing happens that you could not do by clicking. The list is readable without running the page at /webmcp-catalog.json.
Try it “Put a pair of merino socks in my cart, then tell me what is in it.”
This store runs its own MCP server at /mcp. Three tools on it are open to anyone — search the range, read one product in full, read the delivery and returns tables — and three answer only about the customer whose identity the caller can prove. It advertises nothing it cannot serve.
Try it “Which packs under €150 are in stock right now?”
Under the storefront is a real WooCommerce installation: fourteen products with SKUs, stock counts, weights and dimensions, four shipping zones, a customer, five orders, and an open REST API. A merchant connects theirs the same way — the store's own REST keys on the workspace — so the assistant reads the live catalogue instead of a copy of it.
Try it “What is the Quelle 1L made of, and how many are left?”
A demo customer is provided — Mara Oertel, with five real orders in five different states. Ask about an order and a sign-in card appears in the conversation; one click signs you in to this store and returns you to the same conversation, identified. The store signs a short-lived proof with its own key, which the platform checks against the key set this store publishes. No password is shared and no account is created.
Try it “Where has my last order got to?”
When the assistant needs structured details — which order, which item, why it is going back — it mounts a card with real fields and a button inside the chat instead of listing things for you to type. Submitting the card calls the tool straight back.
Try it “I would like to return something from order 53.”
Ask for a human and the conversation is queued to this workspace's Inbox. An operator sees the whole transcript, who the visitor is if they signed in, which tools ran and what each one answered, and can take the conversation over mid-thread — the visitor keeps the same chat window.
Try it “Can I talk to a person about a repair?”
Every layer an agent looks for is served by this store: /llms.txt, /llms-full.txt, /agents.json and its /.well-known twin, /webmcp-catalog.json, a sitemap built from what is really published, and a robots.txt that names each AI crawler instead of leaving them to guess.
Try it “Read this shop's llms.txt and tell me what it can do.”
Every conversation on this shop lands in the workspace Inbox: the full transcript, the tool calls and their answers, and — once a visitor has used the sign-in card — which customer account they are. An operator can reply in the same thread, and the visitor sees it in the same chat window they were already using. Asking for a person is what queues it.
Prefer email? hallo@woo.demo.busymate.ai — a demonstration address on this store's own domain, so there is no real person's inbox in the story.
Reading this as an agent? /llms.txt · /llms-full.txt · /agents.json · /webmcp-catalog.json · /sitemap.xml · /mcp · Photography credits