Markdown export | NookDocs

NookDocs assumes you'll hand pages off to LLMs — to summarise, extend, translate, or regenerate. That handoff needs the source, not the rendered HTML. We expose the source at a stable /path.md URL on every page, plus a one-click clipboard affordance in the UI.

The `.md` URL convention

Append .md to any docs page URL and the server returns the raw MDX source with Content-Type: text/markdown.

# Rendered HTML
https://docs.acme.com/quickstart

# Raw MDX source (same content, different format)
https://docs.acme.com/quickstart.md

Works on any tenant, any custom domain, any path depth:

https://docs.acme.com/api/authentication.md
https://docs.acme.com/configuration/theme.md
https://docs.acme.com/.md                    # root page

Try it on this page: /optimize/markdown-export.md.

What you get back

Exact MDX source from your repo — frontmatter included, components intact, no HTML escaping. Example (abridged):

---
title: Markdown export
description: Every docs page is also available as raw MDX...
icon: file-down
---

NookDocs assumes you'll hand pages off to LLMs...

<ParamField path="...">
  ...
</ParamField>

Headers:

Content-Type: text/markdown; charset=utf-8
Cache-Control: public, max-age=300, s-maxage=300, stale-while-revalidate=3600

No authentication required for public pages. Pages hidden via hidden: true frontmatter return a 404 at the .md URL too.

Use cases

Paste into any LLM

Reader: "summarise this for me"
→ paste /my-page.md contents
→ LLM understands the component set + body + frontmatter

Script-assisted content operations

# Pull a page, run a diff, open an issue when it drifts from your style guide
curl -sL https://docs.acme.com/quickstart.md | ./lint-style.py

# Export every page as MDX for a one-off migration
curl -sL https://docs.acme.com/sitemap.xml | \
  xmllint --xpath "//*[local-name()='loc']/text()" - | \
  xargs -I{} curl -sL "{}.md" -o "./export/$(basename {}).md"

Claude Projects / ChatGPT Custom GPTs

Most LLM providers accept a docs-site URL and will fetch /llms.txt + per-page .md URLs automatically. See llms.txt for the index file format.

Copy from the UI

Every docs page also ships a Copy page split button in the top-end corner of the page header. The primary click copies the raw Markdown to your clipboard; the chevron opens the contextual menu with deep-links to ChatGPT / Claude / Perplexity / Cursor / Windsurf / VS Code / MCP clients.

Click Copy page

Plain MDX source, no decoration. Same content as the .md URL returns.

Click Copy as Markdown for LLMs

Same source prefixed with a component-reference header so the LLM knows the NookDocs component library without fine-tuning.

Click Open in ChatGPT / Claude / Perplexity

Deep-links the provider with the page content pre-loaded as context.

The "Markdown for LLMs" variant wraps the source with an auto-generated component spec (see copy for LLMs for the exact header format).

Limitations

No Accept: text/markdown content negotiation yet. Mintlify lets you send an Accept: text/markdown header to the HTML URL and get Markdown back; NookDocs only exposes the .md suffix for now. The .md URL always works.
No HTTP Link / X-Llms-Txt discovery headers yet. Clients that auto-discover Markdown variants from response headers won't find them. Direct URL access works.
No Cmd+C → full-page-as-Markdown keyboard shortcut. The Copy button is the only UI affordance right now; keyboard shortcut is on the roadmap.
Binary assets excluded. Images and PDFs return a 404 at .md. Only MDX source pages respond.

llms.txt — sitemap-style index of every page's .md URL
Contextual menu — Copy-page dropdown + custom items (ChatGPT / Claude / Cursor deep-links)
Copy for LLMs — the UI variant with component-spec header
SEO — sitemap.xml + robots.txt + canonical URLs

The .md URL convention