Slatesource is built to be read by machines as well as people. Here is how to find everything.
Three entry points, from simplest to most complete.
https://slatesource.com/llms.txthttps://slatesource.com/openapi.jsonhttps://slatesource.com/sitemap.xmlAppend .md or .json to any slate or page URL. Markdown opens with YAML frontmatter carrying the source repository, sync time and content hash, then lists every page of the slate. JSON returns the same page as units, each with its own anchor and URL.
https://slatesource.com/@handle/slate-slug.mdhttps://slatesource.com/@handle/slate-slug/page-name.mdhttps://slatesource.com/@handle/slate-slug/page-name.jsonhttps://slatesource.com/s/69.mdEvery block on a page has a stable anchor. Titles that start with an article, section or chapter number get a short one, so Art. 24 on a page is always #art-24. The anchors are listed in the .json export and linked beside each heading in the .md export.
https://slatesource.com/@handle/slate-slug/page-name#art-24Read access needs no key and no login. Every path below is relative to https://slatesource.com/api and returns JSON. The full request and response shapes are in openapi.json.
GET /slates/{slateId}One slate with its pages and chipsGET /slates/by-slugFind a slate by creator and slugGET /users/{username}A creator profileGET /users/{username}/slatesA creator’s slatesGET /categories/by-pathA category with its slates and creatorsGET /searchSearch slates, creators and categories, plus passages: body-text hits with a snippet, anchor and pathGET /home/discoverThe discovery feedGET /chipsThe chip catalogueA read-only Model Context Protocol server over Streamable HTTP, with no key and no login. Four tools: search finds units by their text, list_pages shows what a slate contains, get_page reads a whole page, and get_unit reads one unit by its anchor, such as a single article.
https://slatesource.com/mcpCreating or editing slates needs an account and a signed-in session. The public spec covers read access only.
Identify your agent in the User-Agent header, cache what you fetch, and please keep to a few requests per second.