The machine view

At a glance

• Add .md to any page address to get its Markdown twin. The home page twin is /index.md.
• Start at /llms.txt for an index of every page.
• Every page also links its twin in its HTML head, as a rel="alternate" link of type text/markdown.
• The site serves the same bytes to every client. It does not change its answer by user agent or by Accept header.

The answer

Fetch llms.txt, pick the pages you need, and fetch each one's Markdown twin. The twin carries the same words as the designed page, with front matter that gives the canonical address, the status, the launch state and the date of last review. Facts the site states are in facts.json.

Files

• Twins. /{page}.md, one for every page. Front matter fields: title, description, url, tier, type, status, launch_state, last_reviewed, index, and facts where a page rests on recorded facts.
• llms.txt. A short index, grouped by purpose, with one line per page.
• llms-full.txt. Every twin in one file, in sitemap order, each introduced by its title and address.
• facts.json. Each published fact with its record id, a value, a source note and a date.
• sitemap.xml. Canonical addresses of the human pages, with the date each was last reviewed.
• robots.txt. The crawl policy and the content signal for search, AI input and AI training.

The machine view in a browser

The Machine option of the switch at the top of every page shows the twin as the page, so a person can see exactly what an agent reads. It is a link, so it works without JavaScript and can be shared. Machine view pages are marked noindex and name the human page as canonical.

Citing

Cite the canonical page address from the twin's url field, with the last_reviewed date. Do not cite a statement as a capability while launch_state is pre_launch or status is planned.

What this does not cover

There is no API, no search endpoint and no MCP server. Nothing on this site needs an account. When product reference exists, it will be added under Reference.