markdownify-mcp: An MCP server that converts PDFs, audio/video, and web pages to Markdown

Based on the MCP protocol, the project provides tools including youtube-to-markdown, pdf-to-markdown, bing-search-to-markdown, webpage-to-markdown, image-to-markdown, audio-to-markdown, docx/xlsx/pptx-to-markdown, and get-markdown-file. Installation uses bun install, with a pre-install step that creates a Python virtual environment and installs markitdown[all]; at runtime, environment variables like MARKITDOWN_PATH, REPOMIX_PATH, and MD_ALLOWED_PATHS override default paths and read boundaries, and MD_SHARE_DIR remains supported as a legacy single-directory alias. Developers using MCP clients like Claude Desktop just add a node + dist/index.js entry to their mcpServers config; teams using the Docker image (mcp/markdownify) need to mount host directories into the container, pass container paths rather than local paths to tools, and restrict read scope with MD_ALLOWED_PATHS. One practical difference: the published Docker image only installs markitdown[pdf], missing the [all] dependencies needed for audio transcription and image OCR, so audio-to-markdown and image-to-markdown fail in the slim image; developers needing full functionality should switch to a local install. The README provides no performance or accuracy data—actual conversion quality depends on markitdown's support for each format.