One tool covers the whole matrix, so the model that just wrote the Markdown also hands you the DOCX you have to send. The reference document is the detail that makes the output usable at work: the headings, fonts and spacing come from a template you supply.
A single-tool server wrapping Pandoc: it takes content inline or a file path, an input format and an output format, and returns the result or writes it to a file.
- Markdown, HTML, plain text and Jupyter notebooks converted between each other and returned straight into the conversation
- DOCX, ODT, PDF, PPTX, RST, LaTeX and EPUB written to a path you name — these formats need an output file rather than an inline reply
- Styling from a reference document for DOCX, ODT and PPTX output, so the result carries your organisation's template instead of Pandoc's defaults
- A Pandoc defaults file in YAML, which turns a set of conversion options into a reusable profile
- Pandoc filters applied during the conversion, for transformations the format options do not cover
- PDF and PPTX as write-only targets — Pandoc has never read a PDF, and PowerPoint input is not offered here
Pandoc installed on the machine — this server drives it and does nothing without it — plus Python with uv. PDF output additionally needs a TeX Live installation. Advanced formats need a complete output path with the filename and extension; without one they may land in the system temp directory. A reference document has to be the same format as the output it styles.
One command — uvx mcp-pandoc