MarkPrep

MarkPrep vs Docling

MarkPrep converts documents to Markdown in your browser, while Docling is an open-source Python toolkit you install on your own computer. Docling analyses PDF layout with AI models and fits into Python pipelines. MarkPrep needs no setup and shows how many tokens the result uses.

Docling reads PDF, Word, PowerPoint, Excel, HTML, EPUB and image files and writes Markdown or JSON. Its code is MIT-licensed, and the AI models it uses carry their own licences.

Side by side

 MarkPrepDocling
Where it runsIn your browserOn your own computer (Python 3.10 or later)
InstallNone, open a web pagepip install docling
Files leave your computerNo, conversion runs in your browserNo, it runs on your own computer
Input formatsPDF, Word .docx, Excel, PowerPoint, CSV, HTML, JSON, XML, Jupyter, EPUB, TXT, images, web pages by URLPDF, Word .docx, PowerPoint, Excel, HTML, EPUB, audio, email, LaTeX, images and more
PDF layout and tablesReads the PDF text layer; flags doubtful tables for you to checkModels for page layout, reading order, table structure and formulas
OCRLocal Tesseract, EnglishBuilt-in OCR for scanned PDFs and images
Token estimatesFor named Claude, GPT and Gemini models, on screenIts HybridChunker counts tokens in Python, with a tokenizer you pick
Context-window checkYes, for the model you pickNo
Chunking for RAGChunk preview split at headings, JSONL exportHierarchicalChunker and HybridChunker in Python
MCP serverYes, in Pro (£19 once)Yes, free (docling-mcp)
Licence and priceFree converter; Pro £19 onceMIT code, free; model licences separate

Where Docling wins

Docling analyses each PDF page for layout, reading order, tables, code and formulas. MarkPrep reads the text layer left to right, top to bottom, and has no maths support. For two-column research papers and pages full of equations, Docling is the better pick.

Docling also reads formats MarkPrep doesn't, such as audio, email and LaTeX. It can read pages with a vision language model such as GraniteDocling. It ships chunkers for retrieval pipelines and a free MCP server, so AI agents can call it directly.

What MarkPrep does differently

MarkPrep needs no Python setup. You open the page, drop in a file and get Markdown, and scanned pages go through English OCR in your browser.

After conversion, the free AI-Ready Pack gives a token estimate for the model you pick and a context-window check. It also shows a chunk preview split at headings and exports those chunks as JSONL. Docling can count tokens and chunk too, but you write Python to see the result.

Frequently asked questions

How do I convert a PDF to Markdown with Docling?+

Install it with pip install docling, then run docling followed by a file path or URL. Docling writes a .md file in the current folder. In Python, the DocumentConverter class does the same job inside your own script.

Can Docling convert Word (DOCX) to Markdown?+

Yes. Docling reads Word .docx files with the same command it uses for a PDF, docling followed by the file path. MarkPrep also converts Word .docx files in your browser and saves the document's images as separate files in a .zip.

Does Docling chunk documents for RAG?+

Yes. In Python code, Docling's HierarchicalChunker makes one chunk for each document element and keeps its headers and captions as metadata. The HybridChunker can then split or merge those chunks so each fits a tokenizer you choose. MarkPrep's chunk preview shows the chunks on screen with no code to write.

Is Docling an IBM product?+

Docling began at IBM Research Zurich, in a team called AI for knowledge. It is now hosted by the LF AI & Data Foundation, under the docling-project organisation on GitHub.

What are good document parsing tools other than Docling?+

Marker and Microsoft's MarkItDown are open-source Python tools that also run on your own computer. LlamaParse is a hosted parsing API that reads files you upload to it. MarkPrep runs in your browser with nothing to install. Each one has its own comparison page on this site.

Can I use Docling and MarkPrep together?+

Yes. Use Docling in a Python pipeline that needs layout analysis, formulas or a free MCP server. Use MarkPrep for a file you want to convert and check without writing code.

Convert a document in your browser and see its token estimate, with no Python to install.