Every documentation page, data model page and forum reading view of Invantive Cloud Documentation is also available as Markdown. An AI agent or a language model then gets the content of the page without the navigation and the layout around it. ## Two Ways to Ask The Markdown of a page is requested in one of two ways: - Append `.md` to the address of the page, such as `/en/docs/invantive-universalsql/invantive-universalsql.md` or `/en/datamodels/exactonline/ExactOnlineREST.CRM.Accounts.md`. This works whatever the request headers say, so a model that cannot set headers can still be pointed at it. - Send the header `Accept: text/markdown`, or prefer `text/markdown` over `text/html` with a quality value, to the normal address. The name of the user agent plays no part. The answer has the content type `text/markdown; charset=utf-8`. The following example reads a table of the Exact Online data model as Markdown: ```text curl -H "Accept: text/markdown" https://documentation.invantive.eu/en/datamodels/exactonline/ExactOnlineREST.CRM.Accounts ``` ## What the Markdown Holds The Markdown of a page starts with a front matter block holding the title, the description, the language and the address of the HTML version under `canonical`, followed by the content: - links to other pages of the site as absolute `.md` addresses, so an agent can keep reading in Markdown, - images as absolute links to the site, - on a data model page, the columns as a Markdown table, - on a platform page, its tables as links, - on a forum page, the question, the accepted answer and the replies, with a link to the topic on the forum. A page that is not yet translated into the language asked for answers in its source language. A request for Markdown never starts a translation. ## Headers Every HTML page names its Markdown version in the header `Link` with `rel="alternate"` and `type="text/markdown"`, so an agent that has the HTML finds the Markdown. Search engines index the HTML only: `robots.txt` asks general crawlers to skip the `.md` addresses, and the sitemaps do not list them. ## `llms.txt` The file `/llms.txt` summarises the site for language models, after [llmstxt.org](https://llmstxt.org/). It is Markdown and holds: - what the site contains, - how the addresses are built, so an agent can reach any page by its pattern, - that every page answers as Markdown, - how to search, also as Markdown, - links to the start pages of the main products, to the data models and to the forum. ## Limits A Markdown request counts as a page request for the rate limits of the site, the same as an HTML request.