Live property search, filters, sorting, currency switching, and map discovery
One connected build
From shop data to a live business experience.
AI works through Selldone MCP, turns approved business data into the experience, then prepares a reviewable launch.
- 01 Connect the shop Scoped Selldone MCP access
- 02 Shape real data Catalogue, products, pricing, and operations
- 03 Build the experience AI creates every responsive surface
- 04 Review and launch GitHub to Cloudflare, with approval
See the complete build logAll 8 verified steps and outcomes
-
01
Connect the shop to Claude
Selldone exposes the shop through an MCP connector. Enabling it gives the AI authenticated access to the same catalogue, pricing, and settings available in the dashboard, so it works with real records instead of invented sample data.
Authenticated access to the live shop catalogue and settings. -
02
Create the repository and clone it
A fresh Git repository is created for the storefront, cloned locally, and opened with a running development server. Everything the AI writes is versioned from the first commit.
A versioned working surface with a local development server. -
03
Fill the catalogue through the backend
Claude creates the category and eight properties directly in Selldone, including titles, descriptions, nightly rates, currency, city, coordinates, room types, and amenities. The records remain editable through the standard dashboard.
Eight real, editable accommodation records in one Selldone shop. -
04
Choose the design reference
Booking.com is used as a concrete visual reference. Its palette, type scale, and spacing are measured and translated into a shared token system for the storefront.
A consistent visual system grounded in a specific reference. -
05
Hand over the screenshots
Each surface is specified with reference captures and a description of its behavior, including what expands, what filters, and what persists between steps. The screenshots become a fast, precise product brief.
Clear visual and interaction requirements for every surface. -
06
Claude builds
The storefront is built surface by surface, with each screen connected to live Selldone data as soon as it appears. Prices, currencies, amenities, and scores come from the backend from the first render.
A complete storefront powered by live shop data rather than stubs. -
07
Test at every width
Real browser screenshots are reviewed at phone, tablet, laptop, and desktop widths after each pass. Visual review catches layout problems that automated checks alone cannot reveal.
Verified responsive behavior across four screen widths. -
08
Go live on Cloudflare
The GitHub repository is connected to Cloudflare so each pushed update can deploy automatically. The storefront is published on its own domain and the live output is compared with the verified local build.
An automatically deployed storefront on a custom domain.