Privacy

Your loan document is handled as the sensitive record it is.

This page describes what actually happens to the file and the figures. It is a description of the system, not legal advice.

The raw file

Uploads are server-mediated. Your browser never receives storage credentials, an object path, or a signed storage URL.
The received bytes are validated for size, declared type, and magic signature before anything is stored.
Storage is private. No policy grants any browser session read, write, or list access to the document bucket.
The raw file is deleted as soon as extraction and confirmation succeed, and in any case on a short timer measured from the moment the upload session is created.
You can delete an analysis yourself at any time, which removes the stored object immediately.

A guest analysis holds the confirmed structured fields and the deterministic result for a short anonymous window, then expires.
If you create an account and save a loan, what is stored is the confirmed structured terms and the calculated result. The raw document is not saved.
Full account numbers, social security numbers, tax identifiers, and bank routing details are never persisted. Where the reader detects such an identifier, it records the category and discards the value.
Deleting your account removes every row it owns and the sign-in identity itself.

Product analytics record which step of the flow was reached, a confidence bucket, a device class, and a generic error code.
They never receive a balance, a rate, a payment, a date, a lender name, a filename, an email address, a loan nickname, or any document text.
Logs never contain document bytes or a full extracted payload. Session replay is off.

Sign-in runs server-side. Authentication tokens never reach your browser and are never written to local storage, session storage, or IndexedDB.
Your session is a rotating, HTTP-only, same-site cookie with a bounded absolute lifetime. Signing out revokes the server session and clears the cookie.
Account data is additionally protected in the database by row level security tied to your own identity.

An OpenAI model reads the uploaded document for extraction only. Requests are sent with storage disabled, and real document processing stays switched off until the operator has verified the applicable data-retention controls.
Official public rate sources are fetched by the server, never by your browser, and each cached value keeps its source, its as-of date, and its retrieval time.
No advertising network, font CDN, or third-party script is loaded on any page of this site.

Operator and privacy contact: [OPERATOR LEGAL ENTITY AND PRIVACY CONTACT TO BE COMPLETED BEFORE PUBLIC LAUNCH].