zcaceres/markdownify-mcp

git clone; bun install (preinstall creates a .venv and installs markitdown[all]); bun run build; bun start
Docker: docker build -t markify-mcp . ; docker run --rm -i -v "$HOME/Documents:/data:ro" -e MD_ALLOWED_PATHS=/data markdownify-mcp

Usage

• •Add to MCP client config: {"mcpServers": {"markdownify": {"command": "node", "args": ["{ABSOLUTE PATH}/dist/index.js"]}}}
• •Available tools: youtube-to-markdown, pdf-to-markdown, bing-search-to-markdown, webpage-to-markdown, image-to-markdown, audio-to-markdown, docx-to-markdown, xlsx-to-markdown, pptx-to-markdown, get-markdown-file (.md/.markdown only)
• •Set MD_ALLOWED_PATHS to a colon (POSIX) / semicolon (Windows) separated list of directories the server is allowed to read; all file-input tools reject paths outside it
• •Override MARKITDOWN_PATH / REPOMIX_PATH if markitdown/repomix are installed system-wide instead of in the bundled venv

Key features

• ✓Converts PDF, images, audio (with transcription), DOCX, XLSX, PPTX to Markdown
• ✓Converts YouTube video transcripts, Bing search results, and general web pages to Markdown
• ✓get-markdown-file tool retrieves existing .md/.markdown files
• ✓Path-restriction security via MD_ALLOWED_PATHS (with deprecated MD_SHARE_DIR alias kept for backward compatibility)

Best for

Giving an LLM client the ability to turn documents, images, audio, and web content into readable Markdown for analysis and summarization.

Caveats

• ⚠The published Docker image installs markitdown[pdf] only — audio transcription and image OCR require the [all] extras and will fail in the slim image (use local bun install for the full feature set)
• ⚠Requires bun plus a Python venv with markitdown; the MCP client config needs the absolute path to dist/index.js
• ⚠In Docker, pass container paths (e.g. /data/foo.pdf) to the tools, not host paths