What Actually Happens When You Upload an .ipynb on Jupy Tools
An honest split: converters process a file for the current request; the viewer and diff stay in your browser. What we do not do, and how to treat secrets.

Privacy copy on converter sites is often wrong in both directions: either “nothing ever leaves your machine” when a download clearly required a server, or “we might keep files” with no detail. Here is the split Jupy Tools actually uses.
Two kinds of tools
Viewer and diff — /ipynb-viewer and /ipynb-diff read the notebook in your browser. You do not need a conversion API to scroll cells or compare two files. Close the tab on a lab PC when you are done.
Converters and most utilities — Word, PDF, HTML, Markdown, ZIP, merge, split, repair, compress, and similar downloads go through our conversion API so we can produce the file you asked for. The upload is processed for that request. We do not keep a notebook library of customer files. Responses use Cache-Control: no-store.
If an older paragraph on the site said a converter “never leaves your laptop,” that was a mistake. This article is the source of truth.
What “processed for this request” means in practice
- You choose a tool and a file.
- The browser sends the bytes (and the options you toggled) to produce the download.
- The server runs the conversion, returns the result, and does not store the notebook as an archive you could log in and retrieve later.
- Guest conversions are credit-limited per day. That limit is about abuse, not about keeping your homework.
We still see the bytes while the request is alive. That is unavoidable for a server conversion. If the notebook is medical, legal, or contains live credentials, do not upload it. Use local Jupyter, nbconvert, or an offline toolchain.
What we do not do
- We do not sell notebook contents.
- We do not train models on your uploads as a product feature.
- We do not require an account for the standard conversion path.
- We do not promise that a shared computer’s browser cache is empty — that is your machine.
Advertising (when enabled) is separate: cookies and ads follow the consent banner and the privacy policy. Ads are not fed your .ipynb.
How to reduce risk anyway
- Strip outputs before sharing: Git hygiene guide.
- Search the notebook for keys and tokens.
- Prefer the viewer when you only need to read.
- Prefer local nbconvert when policy forbids any upload.
- After a conversion on a borrowed laptop, download, then close the tab.
Related reading
- Format choice after you decide an export is allowed: Word vs PDF vs Markdown
- Legal language:
/privacyand/about
FAQ: uploads and privacy
No durable archive. Converter uploads exist to produce that response. Downloads are sent with Cache-Control: no-store. Still treat any internet service as a judgment call for highly sensitive files.
The notebook viewer and the notebook diff read files in your browser. Close the tab on a shared computer when you finish.
No. Strip outputs and secrets first. Conversion is not a vault.

