A PDF or DOCX is read by a parser this page fetches from this site the first time you add one, so you will see a script load. It runs here: your document is opened in this tab and none of its content is sent anywhere.
This is the only thing on this page that sends anything you gave us. When you press it, 10 questions and the text extracted from your documents go to our answering service and on to Anthropic — as text. Your files stay here.
// nothing is sent — check your network tab
Five steps, and only one of them sends anything you gave us.
Worth reading if you are the person who has to sign off on sending a customer's questionnaire to a vendor's website. This is the whole data flow, including the one step that fetches code from us rather than sending anything to us.
The file is read locally
The spreadsheet parser runs in this tab. The file itself is never sent anywhere, in any form.
The analysis is arithmetic
Duplicate clustering and the boilerplate split are deterministic — no model involved, so a visitor who stops here has sent nothing at all.
Your document is opened here
Adding a PDF or DOCX fetches the matching parser from this site — a script, from our origin, so you will see it load. It runs in your tab; no part of the document is sent.
Text goes, files don't
Press the button and ten question texts plus the text extracted from your document go out — used for that request and dropped when it finishes.
The run waits here for you
Answers are held in this browser for seven days so signing up can pick them up. Same device only — they never left it.
The other ninety questions
are the same job.
A trial runs the whole questionnaire against your own Vault — your approved answers, your past RFPs, your voice — and exports back into the original spreadsheet, in place.