Axiomancer

Use these docs with an AI assistant

Every page is published as clean markdown, so you can give an AI assistant the exact, current docs instead of its training memory.

AI assistants answer better when they read the real documentation instead of relying on training-data memory, which may be stale or wrong. Every page on this site is published as clean markdown, so you can hand an assistant the exact, current content for the product you are using.

Three ways to feed the docs to an assistant

One page

Append .md to any page URL to get just that page as markdown. For example, the RouteShift routing guide lives at /routeshift/routing, and /routeshift/routing.md serves the same content as text/markdown.

Two shortcuts live in the header of every page:

  • Copy Markdown puts the page's markdown on your clipboard, ready to paste into a chat.
  • The Open menu alongside it offers View as Markdown plus one-click handoffs that preload the page into Scira AI, ChatGPT, Claude, or Cursor.

The site index

/llms.txt returns a compact index of the whole site — one section per product plus the changelog, one line per page carrying its title, description, and URL. Point an assistant at it first when the job is "find the right page," then have it fetch the specific page it needs.

Everything at once

/llms-full.txt serves the entire corpus — all six products, the changelog, and the utility pages, hundreds of pages at the time of writing — as a single markdown document of roughly 2 MB, on the order of half a million tokens. It is the right move when you want broad context in one shot and the model's context window can hold the full file. Check the response's Content-Length before piping it somewhere expensive.

Working habits that pay off

  1. Scope before you slurp. If you know the product area, open /llms.txt and have the assistant reason over the index first instead of paying for the full corpus.
  2. Prefer the live page over memory. Assistants confidently paraphrase old documentation. Paste the current page or cite its .md URL when the answer must reflect what is deployed today.
  3. Anchor volatile claims. Pricing, limits, and availability change between releases — treat the live page as the source of truth and have the assistant quote from it instead of recalling it.

Example prompts

Read https://docs.axiomancer.io/routeshift/routing.md and explain how to
configure a fallback chain, step by step.
Here is the index at https://docs.axiomancer.io/llms.txt. Which product
should I use to investigate maritime risk, and what page should I read
first?
I read https://docs.axiomancer.io/llms-full.txt. Compare how Overwatch and
Locus each model port congestion, and cite the pages you relied on.

Both endpoints are generated at build time, so they always reflect the current deployment — when docs ship, the markdown ships with them. For volatile facts such as pricing, limits, and availability, treat the live page as the source of truth and confirm before acting on an assistant's summary.

Was this page helpful?

On this page