Manila Support

On your device and in your own private space in your FCT account, so they reach every device you sign in on. That includes the original file or photo you scanned, each page image, and the text recognised from it. Manila requires an account and is not a device-only app; we would rather say that plainly than let the on-device extraction imply otherwise.

Not by ours, and not by a third party. Everyday extraction runs entirely on your device: Apple's text recognition reads the page and on-device foundation models classify it and pull out its fields. Deep Scan, for a long or image-heavy document, is an explicit per-document choice that uses Apple's Private Cloud Compute: Apple's own private infrastructure, never a third-party AI service, and never on by default.

It cannot auto-accept one. Every extracted fact carries the verbatim snippet it came from, and confidence is computed from whether that value can actually be located in the page's recognised text rather than from a model's self-report. A fact with no located evidence never files itself: it comes to you as a card to confirm or reject.

No. Manila asks for write-only calendar access and uses it only when you tap Add to Calendar and confirm the event. It has no ability to read what is already in your calendar.

There is no third-party analytics SDK, no advertising identifier and no cross-app tracking. Manila does report anonymous diagnostics to FCT, crash and performance data, a trail of screen and action names (never content), and feature-use counts, keyed to a random install identifier that is joined to no account and carries nothing you typed, viewed or stored.