llms.txt & markdown endpoints
StyleRef publishes machine-readable surfaces so an LLM can read the product directly. This page lists them.
llms.txt
https://styleref.io/llms.txtAn AI-readable index of the site: a short product description, links to the main product pages, developer and agent tools, example styles, blog posts, the gallery, and key facts. Start here to orient an agent to what StyleRef offers.
The documentation site publishes its own index:
https://docs.styleref.io/llms.txtIt lists every documentation page grouped the way the sidebar is, with a token estimate per page so an agent can budget context before fetching. It is generated from the documentation source, so it never drifts from the pages that exist.
llms-full.txt
https://styleref.io/llms-full.txtThe complete product reference — a few thousand tokens — with example styles, full spec detail, and the pricing table. Use it when an agent needs depth rather than an index.
Raw-markdown endpoints
Content pages are fetchable as raw Markdown, so an agent gets the text without HTML:
| Pattern | Returns |
|---|---|
/blog/{slug}.md | A blog post as Markdown |
/share/{slug}.md | A public style as Markdown — identification header plus the compiled specification |
docs.styleref.io/{path}.md | Any documentation page as Markdown |
curl "https://styleref.io/share/jbnxmk3t-ec81699fbc57.md"
curl "https://docs.styleref.io/introduction/quickstart.md"All of them return Content-Type: text/markdown and an X-Token-Count header so a caller can budget tokens before reading.
Every documentation page has a Markdown twin: append .md to any path on docs.styleref.io. The content is identical; the navigation, sidebar and search payload are gone, which is roughly a third of the tokens.
Styles as Markdown
Append .md to any public /share URL:
curl "https://styleref.io/share/jbnxmk3t-ec81699fbc57.md"You get an identification header — name, description, category, tags, author, palette, remix parent, canonical URL — followed by the compiled specification, ready to paste into any AI tool. About 1,000 tokens, against roughly 28,000 for the HTML page.
For a tool-specific build (flux, midjourney, diffusion, json) or a ready-to-commit STYLE.md, use the REST API or MCP — the .md endpoint always returns the portable default format.
This works only for styles published to the gallery; a link-only (unlisted) style returns 404.
Canonical URLs
Every Markdown endpoint sends a Link: <…>; rel="canonical" header pointing at its HTML page. Markdown can’t carry a <link> tag, so this is how the twin declares that the HTML page — not the .md URL — is the one to index. If you mirror or cache these responses, preserve the canonical.
Style content as data
To fetch a style (rather than a page) as data, use the REST API — it compiles any style to Markdown or JSON and also carries an X-Token-Count header. See Endpoints reference.
Next steps
- Overview — choosing a surface — when to use a static surface versus a live one.
- REST API — fetch styles as data.