Convert Files and Web Content to Markdown

Markdownify is an MCP server that converts various file types and web content to Markdown format, providing tools to transform PDFs, images, audio files, web pages, and more into easily readable and shareable Markdown text. It converts PDF, images, audio (with transcription), DOCX, XLSX, and PPTX files, as well as web content (YouTube video transcripts, Bing search results, general web pages), and can retrieve existing Markdown files.

Use this connector when an AI assistant needs to convert a document, image, audio file, or web page into clean Markdown text for further processing or sharing. When MD_ALLOWED_PATHS is set, every file-input tool rejects paths outside the configured allowed directories, restricting the server's filesystem read access to a known boundary.

Available tools: youtube-to-markdown, pdf-to-markdown, bing-search-to-markdown, webpage-to-markdown, image-to-markdown (with metadata), audio-to-markdown (with transcription), docx-to-markdown, xlsx-to-markdown, pptx-to-markdown, and get-markdown-file (retrieves an existing .md/.markdown file).

Clone the repository, run bun install (whose preinstall step creates a Python virtual environment at .venv and installs markitdown[all]), bun run build, then bun start. For development, bun run dev runs the TypeScript compiler in watch mode; server behavior lives in src/server.ts and tools in src/tools.ts. For a desktop app, configure it with command: node, args: ["{path}/dist/index.js"]. Environment variables: MARKITDOWN_PATH (defaults to the bundled venv's markitdown executable, override if installed system-wide e.g. via pipx install "markitdown[pdf]"), REPOMIX_PATH (path to the repomix executable used by a git-repo-to-markdown tool), MD_ALLOWED_PATHS (a path-delimiter-separated allowlist of readable directories, unrestricted if unset), and MD_SHARE_DIR (a deprecated single-directory alias for MD_ALLOWED_PATHS, still honored for compatibility). For Docker, build with docker build -t markdownify-mcp . and run with a read-only volume mount plus MD_ALLOWED_PATHS set to the mounted container path (not the host path) - the published Docker image (mcp/markdownify) only installs markitdown[pdf], so audio transcription and image OCR (audio-to-markdown, image-to-markdown) require the full [all] extras and will fail in the slim Docker image; use the local bun install setup for the complete feature set. This project is MIT-licensed.

Teams that need to convert PDFs, Office documents, images, audio, or web content into Markdown for AI processing or sharing, with an option to restrict the server's file access to specific directories - especially useful when running in Docker with bind-mounted host directories. Since audio transcription and image OCR depend on the full markitdown extras rather than the minimal PDF-only set shipped in the published Docker image, teams that need those specific capabilities should plan on the local bun install setup from the start rather than discovering the gap only after deploying the containerized version.