Markdown rendering
at engine speed.

Quarterly Report

Revenue grew 34% YoY, driven by API adoption in the enterprise segment.

Code Performance

Inline math: E = mc²

$E = mc^2$

$$
\int_0^\infty e^{-x^2} dx
= \frac{\sqrt{\pi}}{2}
$$

Same parser used by GitLab, Reddit, crates.io

One codebase → browser, server, CLI, edge

190+ languages, server-side applied

LaTeX-quality rendering, no MathJax overhead

Flowcharts, sequences, gantt — SVG output

Your AI chatbot looks like raw text

You built a chatbot with Claude or GPT. It returns Markdown with tables, code, math. Your frontend shows broken formatting or plain text. Users think your product is broken.

// Your chatbot response handler
const stream = await anthropic.messages.stream({ ... })
const markdown = await stream.finalText()

// Before: dangerouslySetInnerHTML with broken formatting
// After: one call
const html = await md.render(markdown)
res.json({ html }) // production-ready HTML

RAG retrieval quality is terrible

You're chunking PDFs and web pages for your RAG pipeline. Raw HTML has noise — navbars, footers, ads, scripts. Your embeddings are polluted. Retrieval precision drops.

// Before: messy HTML chunks with nav, footer, ads
// After: clean Markdown with semantic structure
const markdown = await md.convert(url)

// Split by headings — natural semantic boundaries
const chunks = markdown.split(/^## /gm)
for (const chunk of chunks) {
  await pinecone.upsert(embed(chunk))
}

5 dependencies to render a README

Your docs site needs remark + rehype + shiki + katex + mermaid. Five packages, five version cycles, five configs. Shiki alone is 2MB. A KaTeX update breaks your math. Mermaid conflicts with SSR.

// Before: 5 packages, 200 lines of pipeline config
// import remarkGfm from 'remark-gfm'
// import rehypeShiki from 'rehype-shiki'
// import rehypeKatex from 'rehype-katex'
// ... 15 more imports and plugins

// After: one line
const html = await md.render(content)

Customer sends a PDF, agent can't read it

Your support agent receives PDFs, DOCX files, and URLs from customers. The LLM needs Markdown to reason about them. You're stitching together pdf-parse, mammoth, and cheerio. Each breaks differently.

// Customer uploads a contract PDF
const markdown = await md.convert(file, {
  format: "auto" // detects PDF, DOCX, HTML
})

// Feed to your agent with full structure preserved
const analysis = await agent.run(
  `Analyze this contract:\n\n${markdown}`
)

Obsidian users break your Markdown input

Your app accepts Markdown input. Users paste from Obsidian (wikilinks), Notion (custom blocks), MDX (JSX components), and GitHub (task lists). Half the syntax doesn't render. Users file bugs.

// User pastes Obsidian-flavored Markdown
const input = "See [[Project Plan]] and ~~old text~~"

// Auto-detect flavor, normalize to GFM
const clean = await md.normalize(input, {
  source_flavor: "auto",
  target: "gfm"
})
// => "See [Project Plan](project-plan) and ~~old text~~"

Weekly reports take 2 hours to format

Your team writes weekly reports in Markdown. Converting to PDF for stakeholders means fighting with Pandoc, tweaking LaTeX templates, fixing page breaks. Every week, 2 hours lost.

// GitHub Action: auto-generate PDF on push
const report = fs.readFileSync("reports/week-12.md")
const pdf = await md.render(report, {
  output: "pdf",
  theme: "corporate"
})
await slack.upload(pdf, "#team-reports")

Notion export is a mess of HTML

You're migrating 500 pages from Notion to your new docs platform. Notion's export gives you mangled HTML with inline styles, empty divs, and broken links. Manual cleanup would take weeks.

// Batch convert Notion export
const files = glob("notion-export/**/*.html")

for (const file of files) {
  const html = fs.readFileSync(file)
  const markdown = await md.convert(html, {
    format: "html"
  })
  fs.writeFileSync(file.replace(".html", ".md"), markdown)
}
// 500 pages → clean Markdown in 3 minutes

LLM output renders differently everywhere

The same Markdown from Claude renders differently in your web app, mobile app, email, and Slack bot. Four surfaces, four rendering stacks, four sets of bugs. Users see inconsistencies.

// Same engine, every surface
const markdown = agent.response

// Web app
const webHtml = await md.render(markdown)
// Email
const emailHtml = await md.render(markdown, { theme: "email" })
// Slack
const slackMrkdwn = await md.render(markdown, { output: "slack" })

// Identical rendering logic. Zero drift.