Privacy
What happens to a document you upload, described precisely.
Temporary processing
Converting a document requires reading it, so your file is uploaded to our server and processed there. It is written to a directory named with a cryptographically random identifier, inside a temporary area that is not reachable from the web.
We do not claim that your files are never seen by our systems — they are, because that is how conversion works. What we can state precisely is that they are not stored permanently, are not indexed, are not shared, and are not used to train anything.
Cleanup
An uploaded file is removed at the earliest of these points:
- immediately after it converts successfully;
- when you press Clear, which deletes the whole session;
- when you close the tab, which signals the server to release the session;
- when the session expires — 30 minutes by default;
- when the service restarts, since results are held in memory and leftover directories are swept on startup.
A background worker runs continuously to delete expired sessions, so cleanup does not depend on you coming back.
No account, no history
There is no signup, no login, and no user profile. Nothing links one conversion to another, and there is no document history to browse because none is kept. A session is a random identifier plus a bearer token held in your browser tab; refreshing the page discards it.
What we log
Operational logs record the shape of the work, never its content: format, file size, duration, success or failure, error category, and whether OCR ran. Document text, the Markdown produced from it, and full filenames are never written to logs.
Analytics
There is no third-party analytics, no advertising network, and no cross-site tracking. The only thing stored in your browser is your light/dark theme preference.
Third-party services
In the default configuration, conversion and OCR run entirely on our own server using Microsoft MarkItDown and Tesseract. Your documents are not sent to any third-party API.
The software supports an optional vision-model OCR provider, which would send page images to an external service. It is disabled unless an operator explicitly configures it. If you are running this yourself and enable it, that choice — and the third party’s handling of your data — is yours to disclose.
Open source
Conversion is powered by Microsoft MarkItDown (MIT) and Tesseract OCR (Apache-2.0). MarkPilot is an independent interface built on them and is not affiliated with, endorsed by, or operated by Microsoft.