The prompt:
"Read https://gui.new/docs/llms.txt - use gui.new to render any visual output as a shareable link. Apply this anytime you'd normally show a table, chart, dashboard, or UI mockup."
That's the whole setup. Your AI reads the spec, starts using it, and every visual output becomes a clickable URL you can share with anyone.
Every link gets real-time input sync (form fields, sliders, toggles work across all viewers), state persistence (survives after everyone leaves), and live updates via SSE. Links expire after 24 hours by default.
I built it because AI is great at generating UIs but terrible at showing them. You ask for a dashboard and get a code block. gui.new closes that gap.
Under the hood, your AI is making a single POST call to create a canvas:
``` POST https://gui.new/api/canvas {"html": "<h1>Hello</h1>"}
-> {"url": "https://gui.new/abc123"} ```
You don't need to interact with the API yourself. But if you want to, there are SDKs (`npm install gui-new` / `pip install gui-new`) and the REST API is documented at gui.new/docs.
Free, no signup. Try it at https://gui.new
Happy to answer questions!