Stats
Actions
Tags
From markitdown
Converts binary Office documents (DOCX, PPTX, XLSX), PDFs, images via OCR, HTML, CSV, ZIP archives, and more to structured Markdown preserving headings, tables, lists. Use for formats Claude's Read tool garbles like .docx/.pptx.
How this skill is triggered — by the user, by Claude, or both
Slash command
/markitdown:markitdownThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use `mcp__markitdown__convert_to_markdown` whenever the user needs to read or extract text from binary document formats that Claude's native Read tool cannot parse.
Use mcp__markitdown__convert_to_markdown whenever the user needs to read or extract text from binary document formats that Claude's native Read tool cannot parse.
MarkItDown converts binary and structured document formats into clean, readable Markdown text. It preserves document structure including headings, tables, lists, and other formatting elements.
Supported formats:
Use the markitdown MCP tool when:
The file is a binary Office format: .docx, .pptx, .xlsx, .doc, .ppt, .xls — Claude's Read tool returns garbled binary data for these formats. MarkItDown is the only reliable way to extract their text content.
The user wants structured extraction: When the user needs to preserve table structure, slide content, spreadsheet data, or document formatting in a machine-readable form.
Processing multiple documents in a ZIP: When a ZIP archive contains documents that need to be converted together.
The user explicitly asks to "convert to markdown": Regardless of file type, honor this intent by using MarkItDown.
Do NOT use MarkItDown for:
.txt, .md, .csv, .json, .xml, .html): Use Claude's native Read tool — it handles these perfectly and is faster..py, .js, .ts, .java, etc.): Use the Read tool directly.Call mcp__markitdown__convert_to_markdown with a file:// URI:
mcp__markitdown__convert_to_markdown(uri="file:///absolute/path/to/document.docx")
The tool accepts http:, https:, file:, or data: URIs. For local files, always use the file:// scheme with an absolute path (file:///path/to/file). The tool returns the document content as a Markdown string. Use this output directly to answer the user's question or to further process the content.
When the user asks to read, summarize, or analyze a .docx, .pptx, or .xlsx file:
file:// URI from the absolute path (e.g. file:///home/user/report.docx)mcp__markitdown__convert_to_markdown with the uri argumentDo not attempt to use the Read tool on these binary formats first — it will return unreadable binary data and waste time.
PowerPoint files convert to Markdown with each slide's content. Slide titles become headings, bullet points become list items, and tables are preserved. Speaker notes may also be included.
Excel spreadsheets convert with each sheet's data rendered as Markdown tables. Multiple sheets are separated with sheet name headings.
Word documents convert with full structure: headings, paragraphs, tables, and lists are all preserved in Markdown format.
file:// URIs with absolute paths: Construct the URI as file:///absolute/path/to/file. Relative paths are not supported — resolve to absolute first if needed.uvx markitdown-mcp: It requires uv to be installed on the system. If the tool fails, check that uv is available.npx claudepluginhub pleaseai/claude-code-plugins --plugin markitdownConverts local PDF, DOCX, XLSX, PPTX, images via OCR, and audio files to clean Markdown using Microsoft's markitdown CLI. Best for text extraction from local documents.
Converts PDFs, DOCX, PPTX, XLSX, HTML, images, URLs, CSV, JSON, and more to markdown via tiered fallbacks: MCP markitdown, native tools, or user notice. For ingesting non-plain-text files.
Converts files and URLs to clean Markdown using MarkItDown. Supports PDF, DOCX, XLSX, PPTX, HTML, images (OCR), audio, CSV, and YouTube transcripts. Optimized for LLM ingestion pipelines.