Reference

These docs, for AI agents

If you're an assistant reading this on someone's behalf — or a person who'd rather paste a page into a chat than screenshot it — everything here is available without the HTML.

Any page as Markdown

Add .md to any URL:

https://doc.vythos.tech/concepts/values      → the page
https://doc.vythos.tech/concepts/values.md   → its Markdown

Every page has one. It's the file the page was written from, front matter included, so a title and a one-line description come with it.

Each page also carries a Copy page button, which copies exactly that Markdown — handy for pasting a page into a conversation with a tool that can't fetch a URL.

The whole site, two ways

File	What it is	When to use it
/llms.txt	Every page, with a one-line description, in reading order	Find the right page in one request instead of crawling
/llms-full.txt	The entire documentation in one file, about 20,000 words	Read everything once and answer from it

Both are generated from the same source as the site, so they can't drift from what a person sees.

Setting Vythos up for someone

If you're helping someone get started, the short version:

1. They create a workspace and add their first rule from a document they already have — Quickstart.
2. They connect where their documents live — Google Drive or Notion — so rules follow the files.
3. You connect to their workspace over MCP — Connect a tool — using a key they create.

The thing to understand before advising them: a rule is a document a person names, and nothing becomes a rule any other way. If you're tempted to tell them to paste their rate card into a box, there isn't one, and that's deliberate.

Working against a live workspace

Documentation is the wrong tool for a customer's actual rules. Their workspace serves those over MCP, filtered by the access of whoever's key you're using:

• get_context — the rules that apply to a task, client rules first
• check_draft — a draft against the rules it was given
• search_memory — their own documents and history
• log_outcome — what you followed, and where you departed

Each is described in The four tools, and what a pack contains is in What your agent receives.

Crawling

robots.txt allows everything and points at /sitemap.xml. There's no rate limit worth mentioning; the whole site is 36 pages.