# llms.txt and Markdown export

The build that renders pages for people also projects them for machines. Nothing here is a bolted-on export that can drift: the same content, projected at the same moment, served from the same bundle.

## The directory: /llms.txt

Every page, titled, described, and linked, in the llms.txt convention agents already know:

```bash
curl https://readsmith.dev/docs/llms.txt
```

This site's own output begins, as of Readsmith :

```text
# Readsmith

> The Fair Source documentation platform. Precompiled MDX, a beautiful
> API reference, and agent-ready outputs from the first build.

## Pages
```

Sites with an API reference append an endpoint index, so an agent discovers operations without crawling.

## The full export: /llms-full.txt

The entire site's content in one Markdown document, for tools that want everything in a single fetch. Hidden pages are excluded, exactly as they are everywhere else.

## Every page as Markdown

Each page serves its own Markdown projection under `/md/`: prefix any page's path with it, so this page's projection lives at `/md/ai/agent-outputs`. Operation pages include their generated contract in the projection, so an agent reading the Markdown sees the same parameters a human sees rendered.

## What feeds these

All three derive from each page's built Markdown projection, the same text that feeds search chunks and Ask AI grounding. When you fix a page, every surface updates in the same build; there is no export step to forget.

The capability summary lives one level up: [Agent Skills](/docs/ai/agent-skills) covers `/skill.md` and the `.well-known/skills` convention, which tell agents what they can do with your product rather than where the pages are.