Turn a work order into a job workbook.

One layout has a parser of its own: Invitation Homes. Those are read by our own code, on our own server, and the document text never leaves it. Any other layout has to go through an automatic read, which you approve for that one document.

You get a dialog saying we do not recognise it, and a tick box for that single upload. Ticking it sends the text of that document to our AI provider, Anthropic. A subscription includes 25 of those reads a calendar month and no more than 6 in one day; the free trial includes three. The same allowance covers one other thing: importing a supplier price book whose column names are ones we do not know spends a read too, though only the column headings go out, never the prices you paid. The read only produces a workbook if the items it found add up to a total that is literally printed on your PDF and every item was read with high confidence. Otherwise you get a review table to correct, because a spreadsheet of unverified prices reads as authoritative and should not exist.

That question is exactly why consent is asked per document rather than once at signup. A recognised layout is parsed on our server and its text goes nowhere else. The document is never written to disk on either path. For an unrecognised layout you tick a box confirming you have the right to upload that document and have its contents read, and the answer is never remembered, so the next document gets asked again. You can also switch automatic reading off for the whole account, server side, and no upload can turn it back on.

From an order history you export from your own supplier account and upload as a CSV or spreadsheet. The columns are worked out by what they look like rather than by a per-vendor template, so it does not matter whose export it is. Nothing is scraped from anyone's website, there is no shared pricing database, and the prices in your workbook are the ones you last actually paid. If your export uses column names we do not recognise, working them out spends one automatic read from the same allowance an unrecognised work order comes out of, and what goes out is the column headings and the shape of the file. Never a price you paid.

No. It is a web page, so it has to load over a connection, and there is no offline mode. Once it is open, what you type is held in that browser and syncs back to your account.

$69 a month, and $69 is what the card is charged. The trial is ten work orders over fourteen days with three automatic reads in it, it takes no card, and it does not turn into a paid subscription by itself. Once you subscribe it renews monthly until you cancel, and you cancel in one step from the billing page.

The PDF itself is not. It is processed in memory and never written to disk, on either path, and a recognised layout is parsed on our server and goes nowhere else. Two things are kept, and they are not the same thing. When you build a work order through the review screen after an automatic read, one audit row is written and held for three years: a one-way digest of the file name rather than the name, how many line items you approved, what they added up to, what your document printed as its total, whether those two figures agreed, and whether you ticked the box saying you knew they did not. That row carries no line items, no address and no document. It is the evidence that you saw these numbers before you used them, which is the part of the User Agreement that protects you.

Separately, once you are signed in the job you are working on is saved to your account, and that one does hold the property address and the line items, along with the costs and hours you type in. That is what makes a job open on your phone in the driveway and on the office computer both, and there is no way to have the second without the first. You can download every job you have saved, or delete the lot, from the home page.