qr3.app now serves Markdown — and describes its API machine-readably

More and more requests to a product page do not come from a browser but from a program — a script comparing prices, or an AI assistant someone asked about an integration. Until now both received exactly what a human receives: HTML with navigation, buttons and layout, from which they first have to extract the content.

Two additions change that.

The homepage and the blog are available as Markdown

Ask with Accept: text/markdown and qr3.app answers in Markdown instead of HTML — same content, no scaffolding:

curl -H "Accept: text/markdown" https://qr3.app/en/

This covers the homepage and the blog — the pages that change. The remaining pages are generated once at build time and then served straight from disk; they never reach this negotiation step and stay HTML for now.

Nothing changes for people. A browser still asks for HTML and still gets HTML.

A catalog names the entry points

https://qr3.app/.well-known/api-catalog holds a directory following RFC 9727. It names the REST API, points to its OpenAPI 3.1 description and to the documentation:

curl https://qr3.app/.well-known/api-catalog

The value lies in the fixed address. Anyone building an integration no longer has to hunt for the OpenAPI file in the documentation — their tooling finds it where such directories conventionally live. Alongside it, https://qr3.app/llms.txt describes the developer resources in a form language models can read directly.

What you need to do

Nothing. Existing integrations keep working unchanged; both additions sit alongside what was already there and change no existing behaviour.