Every clip below loops automatically and each one runs the full journey — all the way to a placed order. These are unedited captures of the live agent responding in real time; the product cards, action buttons and confirmations are all real app state.
Lists → set family size to 4 → Generate meal plan → Buy all → Place order.
Starting in Lists, set the household to a family of four and generate a week of meals. The planner returns real recipes — Chicken & Rice Bowls, Penne Pasta Night, Scrambled Eggs & Toast, Cereal & Fruit Breakfast — and a full grocery list scaled to the family: rice, pasta, bread, eggs, milk, vegetables, bananas, coffee and more. One tap adds the whole list to the cart, then straight through checkout to Order placed.
Chat: “Build me a cart for a BBQ weekend for 8 people” → Add all 8 to bag → Place order.
Ask the concierge to cater a BBQ for eight and it scopes quantities, recommends the full cookout set — patties, buns, cheese, condiments, chips, potato salad, soda, charcoal — and offers Add all 8 to bag. From there it flows to the cart and all the way through checkout to a confirmed order.
Home → Visualize with AI → upload your patio → choose a set + lamp → Add all → Place order.
Go into the Home department and scroll to Visualize with AI. Upload a photo of your own patio, pick a complete dining set and an accent piece like a lamp or string lights, and the app stages it all onto your photo with a live “In this look” total. Add all to cart and run through checkout to a placed order.
Upload a coffee-bag photo from the desktop → AI Review Summary → set Coffee on monthly Refill → open Bath Tissue from “Frequently Bought Together” → set it on Refill too → Buy Again shows both on auto-refill.
Upload a photo of a coffee bag straight from the desktop and the storefront’s search-by-photo lands you on the House Blend Coffee page. An AI Review Summary distills what customers say (“Incredible value,” “Great for meal prep”) with a value-for-money gauge. Put the coffee on a monthly Refill & Save, then from the AI-curated “Frequently Bought Together” strip open the Kirkland Bath Tissue and schedule it too. The Buy Again page confirms both items on set-and-forget auto-refill.
Chat: “Sign me up for Gold Star” → then “Upgrade me to Executive” → the Membership page reflects the black Executive card.
The concierge joins the shopper on Gold Star ($65/yr), then upgrades the plan to Executive ($130/yr) — each step with a confirm button right in chat. Navigate to the Membership page and the account already shows the black-and-gold Executive Member card, marked Active. The agent and the page act on one shared account.
Built on Cloudflare — what’s under the hood
The entire storefront and its AI agent deploy as a single Cloudflare Worker — no separate backend to build or operate. Verified in the codebase:
Why it matters: the Agents SDK ships streaming, tool-calling, real-time state, and scheduling out of the box — so this went from idea to a live, globally deployed AI product fast, and it scales from zero to a holiday rush automatically.
Note: the conversational concierge and meal planning are powered by Workers AI (Llama 3.3 70B). The patio “Visualize with AI” feature composites pre-generated product imagery onto the shopper’s photo; swapping in live Workers AI text-to-image generation is a natural next step on the same platform.