Plans that follow your volume

Reading a page costs us real money every time. An unlimited plan would mean pricing everyone for the heaviest user, or quietly throttling the heaviest user. We would rather show you the meter.

One side of one sheet. A double sided form photographed twice is two pages. Pages we fail to read are not charged.

We will not give you a number until we have measured one on your documents. Accuracy on handwriting depends on the hand, the form and the photograph, and a headline figure from somebody else’s paperwork tells you nothing about yours. A pilot answers it properly, on a real sample, before you commit.

It tells you. Every value carries how sure we are and the words it was read from, and anything uncertain is marked for a person rather than passed through. Correcting it takes a tap when our second guess was right.

Per page, quoted from a 50 page sample so you see projected accuracy and cost before committing. Rates fall sharply with volume.

In your workspace, and nowhere else by default. Nobody outside your workspace can see them, every change is recorded against the person who made it, and your documents are never used to train a model. On Enterprise they can stay in your own storage, or the whole product can run inside your network.

Not yet. We will say so here the day we do, with the report available on request. Anyone claiming a certification before the audit has run is telling you something else about how they work.

The content is addressable over an API and a finished document becomes a record with named fields, so it can be posted into anything that accepts one. Named connectors are built with the first customers who need them, which is a conversation rather than a checkbox.

You keep what you extracted. Export everything as spreadsheets and JSON at any time, including the audit trail. There is no format here you cannot take with you.