mdmint

Export Markdown files to DOCX and/or PDF — pure-Python, cross-platform, minimal dependencies.

A single Markdown → HTML pass feeds both writers. The two features that can't be rendered statically in pure Python (mermaid diagrams and LaTeX math) use an optional renderer if one is available and fall back gracefully otherwise — so the base install stays small and works everywhere.

Install

pip install mdmint

The core install is pure-Python (no native libraries): markdown-it-py, mdit-py-plugins, htmldocx/python-docx, xhtml2pdf.

Optional extras:

pip install "mdmint[math]"        # render LaTeX math (adds matplotlib)
pip install "mdmint[weasyprint]"  # higher-fidelity PDF (needs native Pango/Cairo)
pip install "mdmint[all]"         # both of the above

For mermaid diagrams, install the mermaid CLI separately (optional): npm install -g @mermaid-js/mermaid-cli (provides mmdc).

Usage

mdmint notes.md                          # -> notes.pdf
mdmint notes.md -f docx,pdf -o out/      # both formats into ./out/
mdmint a.md b.md -f docx                 # -> a.docx and b.docx (one per input)
mdmint a.md b.md --merge -f pdf -o book.pdf   # one merged PDF
mdmint report.md --toc -f pdf            # insert a table of contents
mdmint report.md -f pdf --pdf-backend weasyprint

| --mkdocs FILE | Build one document from an mkdocs.yml nav (instead of INPUTs). | | --section-style | With --mkdocs: part (default) or heading — see below. | | --template FILE | Word template (.docx/.dotx) for DOCX output. | | --meta KEY=VALUE | Value for a template {{KEY}} placeholder (repeatable). | | --style ROLE=STYLE | Use a specific template style for a role (repeatable). |

Also runnable as python -m mdmint ....

From an mkdocs site

mdmint --mkdocs mkdocs.yml -f docx,pdf -o out/     # -> out/<site_name>.docx/.pdf

The nav decides which pages go in and in what order (without a nav, the docs_dir tree is used, like mkdocs does). Sections become headings:

• --section-style part (default): a top-level section is a part divider that doesn't use up a heading level, so its pages keep their # as chapters; nested sections become #, ##… and demote their pages to match.
• --section-style heading: every section is a heading one level above its pages.

Pages under a section titled Annex(es), Appendix or Appendices become appendix chapters. Links between pages (../other/page.md#anchor) become in-document links; links to pages left out of the nav are kept as plain text. Headings accept mkdocs attr_list syntax (# Title {#id .part}).

Settings can live in the mkdocs.yml itself (command-line options win):

extra:
  mdmint:
    template: templates/brand.docx     # relative to mkdocs.yml
    section_style: part
    status: Final                      # any other key fills {{STATUS}} etc.
    confidentiality: Internal
    styles: {code: "My Code Style"}    # optional role overrides

Branded DOCX with a Word template

mdmint report.md -f docx --template brand.docx --meta status=Draft

With --template, DOCX output is written into a copy of the template: its cover, contents page, headers/footers, page setup and styles are kept, and the content is appended at the end (or where a paragraph reading {{CONTENT}} sits). Text like {{TITLE}}, {{AUTHOR}}, {{DATE}} anywhere in the body, headers or footers is filled from, in increasing priority: mkdocs.yml (site_name → title, site_description → subtitle, site_author → author, extra.mdmint), the Markdown front matter, and --meta. date defaults to the current month; placeholders left without a value are blanked with a warning.

Each element uses the template's styles, picking the first that exists:

How to build a template that styles well — required styles, numbering, placeholders, and a specimen file to test it — is in docs/TEMPLATES.md.

Headings are bookmarked so in-document links work, and the template's TOC field (if any) picks the headings up when Word updates fields on open. Without --template, DOCX output is unchanged (htmldocx).

Library API

from mdmint import convert

convert(["report.md"], formats=["pdf", "docx"], output="out/")
convert(["a.md", "b.md"], formats=["pdf"], output="book.pdf", merge=True)

Supported Markdown

CommonMark plus: GFM tables, task lists, strikethrough, footnotes, auto table of contents, fenced code, admonitions (!!! note / ??? warning callouts), embedded images (relative paths resolved against the source file; remote URLs downloaded), LaTeX math, and mermaid diagrams.

Notes & limitations

• PDF backend. xhtml2pdf (default) is pure-Python and runs anywhere, with a pragmatic subset of CSS. weasyprint produces nicer output but needs native Pango/Cairo libraries, so it is an opt-in extra.
• Math uses matplotlib's mathtext engine (a large LaTeX-math subset, no system LaTeX required) rendered to images. Full LaTeX environments (e.g. align) are not supported. Without matplotlib, math is shown as literal source with a warning.
• Mermaid requires the mmdc CLI. Without it, diagrams are shown as styled code blocks with a warning. A snap-installed mmdc is supported (mdmint places its scratch files where snap confinement can read them).
• Admonitions render as a colour-coded callout box (tinted background + accent bar + coloured title) in both PDF and DOCX. Type icons are intentionally omitted: the pure-Python PDF backend can't embed an emoji font, so colour and the title carry the type instead. The collapsible ???/???+ blocks render fully expanded (static documents have nothing to collapse into).

License

Licensed under the Apache License, Version 2.0.

Copyright 2026 Meaningful Data.