Beautifully typeset PDFs from Markdown.
A real typesetting engine, not a browser pretending to be a printer.
headless browsers involved
Mermaid diagram kinds, drawn as vectors
typical output: selectable, searchable text with embedded fonts
Every document is typeset like a printed book and comes back in one call, the first one and the ten-thousandth.
Quality. What you wrote is what prints. Tables keep their alignment and repeat their headers on every page. Code never spills off the edge. Paragraphs break where a book would break them: no loose lines, no orphans.
Speed. There is no browser to boot, crash, or feed gigabytes of memory. Each document is one call: Markdown in, the finished PDF out.
Reliability. The same call for one document or a whole batch sent from your scripts. Built to be automated, not watched.
Paragraphs set the way books are made (even spacing, clean rag, no loose lines), not the way browsers wrap.
Kerning, ligatures, embedded and subset fonts. Text you can select, search, and copy.
Mermaid charts and TeX math rendered as crisp vectors on the page, not screenshots. Large diagrams get their own perfectly sized page, cross-linked from the text.
A clickable table of contents with page numbers. Bookmarks in the sidebar. Internal links that jump. Task lists that become real, clickable checkboxes.
CommonMark and the GitHub extensions: tables, footnotes, definition lists, alerts, syntax-highlighted code in fourteen languages, color emoji. Front matter becomes document metadata.
Your document and its PDF are gone from our server the moment the PDF is delivered. Images inside the document always render; images linked from the web load for signed-in accounts and are never stored.
Set from plain Markdown. Exactly as the engine returned them.
A realistic multi-page report: metadata, table of contents, metrics tables, task lists, code, footnotes.
Download PDF
TeX math from inline fractions to matrices and piecewise functions, every glyph a vector.
Download PDF
Flowcharts, sequence diagrams, Gantt charts and more, laid out exactly as Mermaid authors expect, on pages sized to fit.
Download PDFThe whole integration:
curl -X POST https://api.legivel.com/v1/render \
-H "content-type: text/markdown" \
--data-binary @report.md \
-o report.pdfTwo ways in.
REST: one call in, one PDF out. Send raw Markdown or JSON with per-document options: themes, feature toggles, light or dark.
MCP: your AI agents write Markdown, and legivel makes it a document. A native MCP server, ready for agent workflows.
More on converting Markdown to PDF, typeset properly.
legivel is in private beta ahead of launch. Tell us what you would generate. We read every request and reply personally.
Thank you, you're on the list. We reply personally, usually within a few days.
Something went wrong on our side. Email us instead: support@legivel.com.
CommonMark plus the GitHub extensions (tables, task lists, footnotes, definition lists, alerts, strikethrough) and beyond: a [TOC] that renders with real page numbers, YAML front matter that becomes PDF metadata, highlighted code in fourteen languages, color emoji.
Yes, as vectors, not embedded screenshots. Thirty-two Mermaid diagram kinds are drawn exactly as Mermaid lays them out; each large diagram gets its own page, sized to the diagram, linked from the text. TeX math is set as vector outlines, inline and display.
Most tools print a webpage to PDF. That is why the output has uneven spacing, tables sliced mid-row, code running off the page. legivel is a typesetting engine: it breaks paragraphs the way books are set, repeats table headers across pages, and keeps widows and orphans off your documents.
That is the point. A REST API for every document, and an MCP server so AI agents can produce finished documents. Everything scriptable, nothing to install.
Images embedded in the document always render. Images linked from the web are loaded too, for signed-in accounts, over a secure connection, each time you convert. We keep nothing: the images, your document and the finished PDF leave our server the moment the PDF is delivered.
It is in private beta. Request access above; beta users shape what launch looks like.