What if your AI chatbot could show you a live dashboard instead of dumping raw HTML into the chat window? That is the pitch behind gui.new, a new tool that acts as a visual rendering layer for ChatGPT, Claude, and other large language models.
The setup is deliberately simple. You paste a single prompt into your AI chat that points the model to gui.new's documentation. From that point on, whenever you ask for something visual - a chart, a form, a dashboard, a UI mockup - the model renders it as a live, shareable link instead of outputting code blocks you would need to copy and run yourself.
How It Works
The prompt instructs the AI to use gui.new's API whenever it would normally generate visual output like tables, charts, or interface designs. Instead of writing HTML or React code inline, the model sends the output to gui.new's servers, which render it and return a URL. You click the link and see a fully interactive result.
This solves a genuine annoyance. Anyone who has asked ChatGPT to "make me a dashboard" knows the result: a wall of code you need to paste into a file, open in a browser, debug, and iterate on. Gui.new collapses those steps into a single click.
The shareable link angle is useful too. Instead of screenshotting a chart or sending someone a code snippet, you send them a URL they can interact with directly.
Limitations to Consider
The tool depends on the AI model correctly following the gui.new prompt instructions every time. In practice, models sometimes ignore system prompts or revert to their default behavior of dumping code. The rendering also happens on gui.new's servers, so there are questions about persistence (how long do links stay live?) and privacy (what data passes through their infrastructure).
For quick prototyping and internal demos, this looks genuinely useful. For anything production-facing or sensitive, you would want to understand the data handling before relying on it.