MarkPrep vs Microsoft MarkItDown
Microsoft MarkItDown is a well-regarded, MIT-licensed Python library that converts a wide range of documents to Markdown using text-layer extraction. MarkPrep aims at the same open-source quality tier for text-based documents, but it runs entirely in your browser and adds the AI-ready tooling MarkItDown was never meant to include. If you want a scriptable Python library, MarkItDown is excellent. If you want zero install, guaranteed privacy, and output that is ready for an LLM, MarkPrep is the faster path.
The honest framing: these two are not competing on raw conversion quality. Both use text-layer extraction and both are weaker than GPU deep-learning pipelines on scanned pages and dense tables. The real difference is everything around the conversion — where it runs, what you have to install, and whether the result is shaped for retrieval and prompting.
Side by side
| MarkPrep | MarkItDown | |
|---|---|---|
| Where it runs | Browser (WebAssembly), local | Python library on your machine |
| Install required | None — open a web page | Python + dependencies |
| Files leave your device | Never | Never (you run it locally) |
| Token counting | Live, per-model | None |
| Context-window verdict | Yes | None |
| Chunking for RAG | Heading-aware preview + JSONL | None |
| OCR | Local Tesseract, offline | None built in |
| Format coverage | PDF, DOCX, XLSX, PPTX, HTML, CSV, JSON, XML, EPUB, IPYNB, TXT, images, URL | Broad |
| License / price | Free tier unlimited; Pro $4/mo | MIT, free |
When to use which
Use MarkItDown when you are already working in Python, want to embed conversion inside a script or pipeline, and do not need token counts or chunking. Its MIT license and broad format support make it a dependable building block.
Use MarkPrep when you do not want to install anything, when the files are sensitive and must stay on your device, or when the whole point is to feed an LLM. The live token count, context-window verdict, heading-aware chunk preview, and JSONL export with heading-path and page metadata turn a raw Markdown dump into something you can drop straight into a RAG pipeline.
Frequently asked questions
Is MarkPrep just MarkItDown in a browser?+
Not quite, but they aim at the same quality tier. MarkItDown is Microsoft's open-source Python library for turning documents into Markdown. MarkPrep targets parity with that text-layer quality while running entirely in your browser via WebAssembly, so there is no Python environment to install and nothing leaves your machine.
Does MarkItDown count tokens or chunk for RAG?+
No. MarkItDown converts documents to Markdown and stops there. It has no token counting, no context-window awareness, and no chunking. MarkPrep adds a live token count, a context-window verdict, heading-aware chunk preview, and JSONL export with heading-path and page metadata.
Which handles PDF tables better?+
Both rely on text-layer extraction, so both are weaker on complex PDF tables than GPU deep-learning tools. MarkPrep flags questionable tables in its Verify tab so you can spot them, but neither MarkPrep nor MarkItDown matches a GPU pipeline on hard tables or scans.
Do I need to install anything to use MarkPrep?+
No. MarkPrep is a web page. MarkItDown is a Python library, so running it yourself means setting up Python and its dependencies. MarkPrep runs in the tab you already have open.
Is MarkItDown open source?+
Yes, MarkItDown is MIT-licensed on GitHub with broad format coverage. If you are comfortable in Python and want a scriptable library, it is a solid choice. If you want zero install, privacy, and AI-ready output, MarkPrep is built for that.
Try it in your browser
No install, no upload, no account. Convert a file and see the token count in seconds.