Early access—73 of 128
scrollwork
Boceto — Jewellery atelier: the hero is the gouache drawing of a ring and the cursor is the brush; wherever you paint, the real piece appears, and then the stone, the pavé and the gold are named.

What to expect

Every generation is different: your AI makes its own version, not a pixel copy. What the prompt pins down is the structure, the rhythm and the motion, with their numbers and acceptance criteria; what changes are the visual details, the images and the copy you put in.

To make it yours: swap the copy (4 written blocks), the palette (4 colours with their use) and the photos for yours, at the same ratio. The structure and the motion stay.

More like this

Miga — Sourdough bakery: the cursor is the blade that scores the raw dough, and as you scroll the loaf bakes from the crust inwards until its ear opens.
New
Premium
Miga
Nocturno — Photographer portfolio with keyboard navigation and a gallery that behaves like a film reel.
Premium
Nocturno
Vatio — Energy drink: a 3D carousel of eight cans under a halo; on scroll the chosen can grows, spins and lights up each ingredient the copy mentions on its back.
New
Premium
Vatio
Plancha — A late-night burger joint: the hero photo turns cream, the burger splits into five labelled layers, then drops into its combo box with fries, sauce and shake.
New
Premium
Plancha

How to use it

From prompt to a live page

  1. 01

    Copy the prompt or the source code

    The prompt carries the whole page: structure, motion and the stack notes the model needs. It's written in Spanish and asks your agent to build all visible copy in English. With Full Stack you can also download its source code.

  2. 02

    Paste it into your AI

    Written and checked with Claude Opus 5, and the Obsidiana prompt has also been tested with ChatGPT. Any agent that accepts a long prompt can use them, but we haven't measured them against others.

  3. 03

    Make the content yours before polishing

    Change the copy, the colour tokens and the images first. Tuning the motion before you have the real content is wasted time.

  4. 04

    Bring it into your project

    Drop it in as a self-contained page, or ask your agent to rebuild it in your framework. The prompt isn't tied to any stack.

FAQ

It's a prompt: structured language that describes the piece precisely enough for your AI to rebuild it. With the Full Stack plan you also get the ready-made source code.

Yes. By default it generates a self-contained HTML page; add one line and your agent builds it in Next.js, Astro, Vue or WordPress. There's no framework dependency because there's no code tied to any.

All of it. The structure and the motion are what you take; the theme, the colours and the copy are yours. Changing it is literally asking your AI in the same conversation.

Yes. With any paid plan: a commercial license for your own and your clients' projects, no attribution and no limit on sites. It doesn't allow reselling or redistributing the prompts or the code as they are. The free layers are for trying the product: turn them into something of your own rather than publishing them untouched.

On a regular basis. With the one-time payment, everything added after your purchase is yours too; with the yearly plan, whatever is added while your subscription is active.

It won't be 100%. Every generation is different: your AI makes its own version, not a pixel copy. What the prompt pins down is the structure, the rhythm and the motion, with their numbers and acceptance criteria; what changes are the visual details, the images and the copy you put in. If you need the exact piece, the Full Stack plan includes its source code.

Want it in your framework?

Add one line at the end of the prompt and your agent builds it in Next.js, Astro or whatever you use, with your own design tokens.

How to do itConnect the MCP