---
title: "llms.txt and markdown endpoints"
url: "/help/buuk/llms-txt-and-markdown"
updated: "2026-10-22T11:13:49.734Z"
---
# llms.txt and markdown endpoints

AI assistants and agents do not always browse HTML well. They read plain text and markdown. Your help center carries a parallel surface for them: an `llms.txt` index, per-section lists, and a markdown mirror of every published article and API operation.

The URLs are predictable, so a tool can discover a page, fetch a clean version of it, and cite the source — the full pipeline without converters.

## llms.txt

The index lives at `/help/your-slug/llms.txt`. It is plain text with a stable shape:

```
# your-slug Help Center
> This file lists all publicly available help center articles for use by AI agents and language models.
> Source: https://app.getbuuk.com/help/your-slug

## Help center
- [Introduction](/help/your-slug/introduction): /help/your-slug/introduction/content.md
```

- Published help articles grouped by section, each with the page link paired to its `content.md` mirror.
- A `Sections` list pointing at one llms file per section.
- A heading per workspace tab, listing the pages inside — including guides and FAQ links.
- An `API Reference` block: each spec, its version, and every operation listed as `METHOD /path — Summary`.

Add `?lang=` with a published locale and the titles and section names come back in that language. The file is cached for an hour, so edits reach AI tools on their next fetch.

## Per-section llms files

`/help/your-slug/llms/section-slug.txt` lists the published articles of one section, in the same title-and-mirror format. Point a tool at one file when it only needs that topic.

## The markdown mirror

Append `content.md` to any article URL — `/help/your-slug/some-article/content.md`. The response is the article as markdown, headings, tables, and code intact, with reusable content blocks expanded. If the reader's locale has a published translation, that is what the file serves.

API reference works the same way: `/help/your-slug/reference/operation-id/content.md` returns one operation as flattened markdown, matching the human page's access rules and version pinning. Drafts, private sections, and pages restricted to the AI assistant never return a file.

## Who uses this

- AI assistants and chat tools that summarize docs and cite the source page
- Coding agents and MCP-based tools that fetch docs mid-task
- Search pipelines that index docs without parsing HTML

This is generative engine optimization in practice: when a tool gets clean markdown instead of scraped HTML, your pages are more likely to be quoted — and cited. **Doc health** tracks how consumable your docs are for AI tools as Agent readiness, so you can watch the score move.

The changelog has its own feed at `/help/your-slug/changelog/rss.xml`, so release notes reach feed readers and integrations the same way, with tags as categories.

## Access controls

These endpoints respect help-center visibility. A private help center answers **Not available** for llms.txt, per-section lists, and content.md — a tool without membership gets blocked at the same gate as a human visitor.

On Growth with a custom domain, llms.txt builds its links from that public origin when one is configured, so the entries stay valid for readers and tools alike.

## Related

- Ranking in classic search: [SEO metadata and social images](./seo-metadata-and-social-images).
- Crawlability: [Sitemap and robots](./sitemap-and-robots).
- How AI readiness is scored: [Doc health and freshness](./doc-health-and-freshness).
- Questions from readers, answered in place: [MCP for readers](./mcp-for-readers).
- Gated help centers at a glance: [Private help centre](./private-help-centre).