MarkItDown — convert files and documents to Markdown

The problem it solves

Real-world documents live in PDFs, Office files and images that language models and text pipelines cannot read directly. Converting them into clean, structured text is a tedious, error-prone step.

What is it?

MarkItDown takes many common file formats and outputs Markdown, preserving structure like headings, lists and tables where possible. Because Markdown is compact and text-based, the result is well suited for LLM ingestion, RAG and indexing.

Why it's getting attention

With roughly 163k GitHub stars and Microsoft's backing, it rode the wave of RAG/LLM tooling. As more teams build document-aware AI, a reliable file-to-Markdown step became a common need.

Key features

• ✓Converts PDF, Word, PowerPoint, Excel, images and more to Markdown
• ✓Preserves structure (headings, lists, tables) where possible
• ✓Python library plus command-line usage
• ✓Designed for LLM/RAG ingestion
• ✓Maintained by Microsoft

Best use cases

• •Prepare documents for RAG or LLM context windows
• •Bulk-convert an archive of Office files to Markdown
• •Build a document-ingestion step in an AI pipeline
• •Normalize mixed file types into one text format for indexing

How to install / try

MarkItDown is a Python package installed with pip. See the repository for the exact package name, optional extras for specific file types, and current usage.

How to use

Use it from the command line to convert a file to Markdown, or import it as a Python library to convert files programmatically inside a pipeline. Refer to the repo for supported formats and options.

Alternatives

Who should try it — and who should skip

Try it if you build LLM/RAG features and need documents as clean Markdown. Skip it if you need high-fidelity human-facing document conversion or strict layout preservation.

Frequently asked questions

Related repositories