About Licenses when using markitdown-ocr

[k-kondohtks]

When I installed markitdown-ocr, "pymupdf" was also installed at the same time.
I believe "pymupdf" is under the Affero GPL 3.0 license, but is it possible to use "markitdown-ocr" under the MIT License
when there are libraries including Affero GPL 3.0 as dependencies?
I am concerned about whether there would be any licensing issues when using it for a commercial service.

[pauldev-hub]

Hey @k-kondohtks ,

Your concern is valid and technically correct. Here's the full picture.

The core issue — AGPL is a strong copyleft license
You cannot deploy PyMuPDF (AGPL) as part of a server-based application or service without disclosing your own application's full source code under AGPL to any users interacting with it. If you can't meet those requirements, a commercial license is required. (Artifex)

So yes — even though markitdown-ocr itself is MIT, the fact that it pulls in pymupdf as a dependency means your entire commercial service becomes subject to AGPL obligations when you use the OCR feature.

MIT + AGPL dependency = the MIT license doesn't protect you

AGPL imposes strong copyleft obligations that may require open-sourcing the entire application — which runs counter to the permissiveness of the MIT license. (GitHub) The MIT license of markitdown-ocr only covers markitdown's own code. It cannot override pymupdf's AGPL terms.

Your two options for commercial use

Option 1 — Purchase a commercial PyMuPDF license from Artifex
PyMuPDF is available under AGPL for open-source/freeware use, or a commercial license for proprietary use without AGPL obligations. (Pymupdf) Contact Artifex at artifex.com/licensing to get a commercial license — this removes all AGPL restrictions.

Option 2 — Use markitdown without the OCR extra
Install without pymupdf dependency:

pip install markitdown # no OCR extra

instead of:

pip install markitdown[ocr] # this pulls pymupdf

Then use an alternative OCR library with a permissive license (e.g. pytesseract under Apache 2.0) if you need OCR functionality.

Summary verdict
Using markitdown-ocr in a closed-source commercial service as-is violates AGPL unless you either open-source your entire service or purchase a commercial PyMuPDF license from Artifex.

⚠️ Note: I'm not a lawyer — for definitive legal advice on your specific deployment, consult a legal professional.

👍 If this helped you, please mark it as the answer — it helps others in the community who run into the same issue find the solution faster!

[k-kondohtks]

Hi @pauldev-hub,
Thank you for your answer and the two types of proposals for commercial use.
I will take your suggestions into consideration when actually using it commercially.

Once again, I would like to express my gratitude.
k-kondohtks