Summary
Everything arrives as markdown, whatever it started as.
What it takes off you is the per-format handling. A PDF, a deck, a spreadsheet and a video transcript all come back in the same shape, so a mixed pile of sources can be read in one pass instead of one converter at a time.
What it is
A converter with one output format. Each call takes a file path or a URL and returns markdown — including a transcript for audio and video, and a description for images.
What you get
- Office files converted: DOCX, PPTX and XLSX
- PDFs converted to markdown
- Audio transcribed where transcription is possible, and a YouTube video returned with its transcript where one exists
- An image returned as its metadata and a description
- A webpage, or a Bing results page, converted to markdown
- A whole git repository flattened into one document — file tree and source — from a GitHub URL or an owner/repo shorthand
- A markdown file already on disk read back by absolute path
Requirements
Npx on your PATH. The file calls take an absolute path on the machine the server runs on; the URL calls need network access.
Setup effort
One command — npx -y mcp-markdownify-server