Skip to content
Markdown

Responsible use

Learn from the web. Make something of your own.

Designers have always studied the work they admire. Markdown makes that study faster and more structured — while keeping a clear line between inspiration and imitation.

What we never extract

  • Logos inside generated systems or exports
  • Illustrations and icons
  • Photography and video
  • Written content and copy
  • Proprietary font files
  • Source code and scripts
  • Brand names presented as your own

How to use results well

  • Treat every system as a starting point, then adapt it to your own brand and content.
  • Keep the attribution that ships with every export.
  • When you customise a style, move the “How different?” slider to at least “Clearly different” for commercial work.
  • Write your own copy and commission or create your own imagery.
  • License any typeface you use; Markdown only recommends open-licence fonts.

How analysis works

When you submit an address, our server requests that single public page and up to six of its stylesheets. We read CSS values — colours, font-family names, sizes, radii, shadows, spacing, transitions and container widths — and count how often they occur.

Those observations are reduced to coarse signals (for example “a warm off-white background, a serif display face, soft 12px corners”). A new system is then composed from those signals: we build our own type scale, spacing scale, semantic palette, component rules and dark mode. The originality control shifts hues, swaps typefaces and reshapes geometry further.

Proprietary or system fonts are always replaced with open-licence alternatives. Nothing from the page’s markup, text, images or scripts is kept, and the page itself is not stored on our servers. Your projects live only in your browser.

On curated studies we show the source’s logo next to its name, purely so it’s clear whose work is being studied. Logos never appear inside a generated system, its previews or its exports, and they remain their owners’ trademarks.

If a page can’t be read (for example, it relies on JavaScript or blocks automated requests), you’ll receive a clearly labelled estimated system instead.

Safe requests

  • Only public http/https addresses on standard ports are accepted.
  • Private, loopback, link-local and internal network addresses are refused — including when a domain resolves to one, or redirects to one.
  • Requests have strict time and size limits, follow at most four redirects and are rate-limited.
  • We honour robots.txt rules for our crawler and for all agents.

For site owners

Our crawler identifies itself as MarkdownStyleBot/0.1. To opt out, add the following to your robots.txt, and we’ll decline to analyse your pages:

User-agent: MarkdownStyleBot
Disallow: /

Attribution to your site is preserved on every generated system and export.

Browse sites