A versatile Microsoft-built utility for converting diverse document formats, images, and audio into structured Markdown for LLM analysis.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install ms-markitdown
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install ms-markitdown using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is MarkItDown?
MarkItDown is a lightweight Python utility developed by Microsoft specifically designed to bridge the gap between legacy document formats and modern AI-driven text analysis. It streamlines the conversion of various file types—including PDFs, Word documents, Excel spreadsheets, and even YouTube videos—into clean, structured Markdown. This makes it an essential tool for developers and researchers who need to prepare data for RAG (Retrieval-Augmented Generation) pipelines or provide high-quality context to agents within the Openclaw Skills ecosystem.
Beyond simple text extraction, MarkItDown intelligently handles complex document structures. It can preserve table formatting, process images via OCR to extract text from screenshots, and even transcribe audio files. By standardizing diverse inputs into a single, LLM-friendly format, it significantly reduces the friction of document ingestion and data cleaning tasks.
MarkItDown Use Cases
- Converting multi-page PDF research papers into structured Markdown for AI summarization.
- Extracting text and metadata from image-based screenshots using built-in OCR capabilities.
- Transcribing audio recordings or YouTube videos directly into text for searchable archives.
- Automated batch conversion of legacy Word and PowerPoint files into clean documentation.
- Integrating document-to-text conversion into automated workflows using Openclaw Skills and the Model Context Protocol.
How MarkItDown Works
- The user specifies a source file, directory, or URL (such as a YouTube link) to be processed by the MarkItDown utility.
- The tool identifies the file format and routes it through specialized converters, utilizing libraries like Tika for documents or OCR engines for images.
- The engine extracts the raw content while attempting to preserve structural elements like headings, lists, and tables.
- Audio files and video URLs are passed through transcription layers to convert speech into text blocks.
- The final output is synthesized into a standardized Markdown format, ready to be utilized by other Openclaw Skills or saved to a local file.
MarkItDown Setup
To get started with MarkItDown, ensure you have Python 3.10+ and Java 11+ installed. Use pipx for a clean, isolated installation of the full suite:
pipx install 'markitdown[all]'
For basic document support (PDF, Word, PPTX) without audio or OCR dependencies, you can use a minimal install:
pip install 'markitdown[pdf,docx,pptx]'
MarkItDown Data Schema & Taxonomy
MarkItDown organizes data by mapping source document elements into Markdown syntax. The schema follows a predictable transformation pattern to ensure consistency across different input types:
| Source Element | Markdown Mapping | Notes |
|---|---|---|
| Headings | #, ##, ### | Preserves hierarchical structure |
| Tables | GFM Tables | Formats Excel and Word tables into Markdown |
| Images | Extracts EXIF data and performs OCR for alt-text | |
| Lists | - or 1. | Maintains bulleted and numbered sequences |
| Hyperlinks | Text | Preserves internal and external document links |
| Metadata | YAML/Text | Captures file properties and document info |
MarkItDown Advanced Features
- MCP Server Integration: Deploy as a Model Context Protocol server for seamless interaction with LLM applications.
- YouTube Extraction: Provide a video URL to automatically fetch and transcribe content into a Markdown file.
- Archive Support: Recursively process and convert files contained within ZIP archives.
- Python API: Use the
MarkItDown()class within your own scripts for programmatic document processing in Openclaw Skills. - Fine-grained PDF Control: Use the
--pagesflag to extract specific content from large, multi-hundred-page documents.
SKILL.md
Loading
Related Openclaw Skills
DocsAgent
A local-first documentation engine for performing hybrid semantic and text searches across private PDF, PPTX, and DOCX files.
Clock Video Production
An AI-powered skill that converts clock product imagery into polished 1080p promotional videos using cloud GPU rendering.
Travel Allowance Assistant (fb-travel-allowance-skill)
An intelligent travel allowance assistant that automates expense calculations, policy compliance, and approval workflows for corporate business trips.
Intelligent Expense Assistant (fb-expense-skill)
An AI-driven reimbursement assistant that automates invoice recognition, travel matching, and multi-level expense approval workflows.
Program Video — Sequence and Export Edited Videos
A cloud-based AI video editing skill that sequences raw footage into professional 1080p MP4 programs using GPU-accelerated rendering.
Patch Maker
A professional AI-driven video repair skill that patches out unwanted regions and exports clean, high-definition footage via cloud GPU rendering.
Featured*
