Markdown without drift

The feature#

Every route projects a markdown surface: GET /why.md, GET /docs/<slug>.md. For content-backed pages the projection serves original — the file you wrote, byte for byte, frontmatter included:

content/posts/*.md  →  june gen  →  app/_content.ts  →  POSTS / DOCS
                                         │
        HTML view ◄──── one manifest ────┼──── .md projection (original, verbatim)
                                         └──── search_site / get_page MCP tools

Dev and the built worker read the SAME frozen manifest, so there is no "works locally, differs deployed" for content. Most frameworks reconstruct markdown from rendered HTML; June serves the source, so a diff against your repo is empty.

Linkable headings#

The HTML view gives every heading an id, with GitHub's rules, so a link you'd write on GitHub works here too:

## Q & A           →  <h2 id="q--a">Q &amp; A</h2>
### Use `ctx.user`  →  <h3 id="use-ctxuser">Use <code>ctx.user</code></h3>
## 邊緣排版          →  <h2 id="邊緣排版">邊緣排版</h2>

• The rules: lowercase; punctuation, symbols, and emoji drop; letters of every script stay; each space becomes -; a repeated heading gets -1, -2 in document order. A heading with nothing left becomes section.
• Raw HTML headings count too. A heading you write as HTML keeps its attributes. If you gave it an id, that id is kept as your explicit anchor. A generated id never repeats one already on the page.
• Only the id. June adds no link or class. How an anchor looks is your app's decision; this site adds its # in the page component.
• entry.headings lists each heading as { depth, text, id }, in order. Use it for a table of contents (the "On this page" rail here), or put it in a .json projection so an agent gets the page's structure and a deep link to every section.
• One algorithm, shared. Code that needs the same ids, such as a search index built from the markdown source, imports it rather than re-implementing:

import { createSlugger, slugify } from "@junejs/core/slug";

const slug = createSlugger(); // one per document, called in heading order
slug("Setup"); // "setup"
slug("Setup"); // "setup-1"

The .md projection is unchanged: it's still your file, with no ids added.

Try it on this page#

curl -s https://june.build/docs/features-markdown.md

That response IS this file in our repo — the same bytes git show would give you. Our site tests assert it with strict equality, not a contains.

Why it matters#

Markdown is the densest, least ambiguous surface you can serve an agent. Serving the authored source means what an agent reads is exactly what you wrote — and frontmatter arrives as structured metadata instead of being boiled away by a renderer.