MarkItDown
microsoft/markitdownReviewed by AI
What it does
MarkItDown is a Microsoft Python tool that converts files into Markdown, a simple text format that AI chatbots and text-analysis tools read well. It handles PDF, Word, PowerPoint, Excel, images, audio, HTML, CSV, JSON, ZIP, EPUB and YouTube links, and keeps headings, lists and tables. The README says the output is meant for software, not for faithful copies for people to read. You need Python 3.10 to 3.14 and use it from the command line or in Python code. Optional Azure and OpenAI features are paid.
How easy it is to install
C · install 1 or 2 tools · Windows
- Windows C
- macOS C
- Linux C
The work is the same on all three: install Python, run pip install 'markitdown[all]' and then type a command such as markitdown file.pdf -o document.md for each file. The README's virtual-environment commands are written for Mac and Linux shells.
Numbers on GitHub
- Stars
- 193,246on Oct 31, 2026
- Downloads
- —all releases
- Growth
- +864stars in 5 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Python
- To build from source
- Docker
- Disk space
- not determined
- Accounts or keys
- Azure Document Intelligence (optional cloud conversion), Azure Content Understanding (optional cloud conversion, audio and video), OpenAI or a compatible service (optional image descriptions and OCR plugin)
- Cost
- Free · Free under the MIT license; built-in conversion runs on your computer. Azure Document Intelligence, Azure Content Understanding and OpenAI image descriptions are optional and billed by those services.
- License
- MIT (Permissive)
- Can replace
- Mathpix
- Last update
- Oct 22, 2026
Risk signals (3)
- Infocan fetch remote urls
README.md: "MarkItDown performs I/O with the privileges of the current process" and "convert() method is intentionally permissive and can handle local files, remote URIs, and byte streams." The README asks developers to sanitize untrusted input.
AI review
- Infothird party plugins
README.md: "MarkItDown also supports 3rd-party plugins. Plugins are disabled by default." Turning them on with --use-plugins runs code from other authors.
AI review
- Infocloud options send documents
README.md comparison table: Azure Document Intelligence and Content Understanding are "Cloud layout extraction" / "Cloud multimodal extraction" with "Billable Azure API calls"; the OCR plugin uses "LLM Vision" through an OpenAI client. With these options your files leave your computer.
AI review
Evidence is shown in its original language.
What Radar could not check
- How good the conversion is for complex layouts, scanned PDFs and large spreadsheets without the cloud options.
- Exact Windows steps: the README shows Mac/Linux commands such as source .venv/bin/activate.
- Whether audio transcription needs ffmpeg installed separately on Windows.
- What third-party plugins found through #markitdown-plugin actually do.
- Real cost of the Azure and OpenAI options per document.