Questions, answered.

Yes. The .mbox file you drop in stays in your browser. Parsing, classification, and the dashboard all run in a Web Worker on your device — there's no server endpoint to upload to. You can verify this by opening your browser's Network tab during a scan and watching it stay quiet. The Why-local section above goes deeper.

It's a standard mailbox archive format. For Gmail, use Google Takeout (Mail → mbox). For Apple Mail, File → Export Mailbox. For Thunderbird and most desktop clients, your account folder is already an .mbox. The export-instructions section on the home page walks through the major providers.

Yes. Detection patterns cover English, Spanish, French, German, Italian, Portuguese, and a chunk of Catalan. The OCR layer ships with traineddata for English / Spanish / French / German. Other languages still parse but may classify less reliably until we add their cues.

Because adding it would mean your email could go to our servers — and that's the line we don't want to cross. The point of local-scan is that we don't ever have your mailbox. We may explore an OAuth path later if it can run entirely in-browser without a server-side relay.

No. sub-hunt is email-based — it scans for the receipts your services email you. If you'd like a bank-statement equivalent, our recommendation is to run sub-hunt to get the full subscription list, then verify against your statement manually.

It's a heuristic, not a ledger. Real subscription receipts usually surface with the right brand, amount, and cadence; tricky cases (Google Play receipts that bundle YouTube Premium, multi-locale Stripe disclaimers, marketing emails dressed up as invoices) sometimes get the brand or kind wrong. Every finding shows confidence and lets you view the source email so you can verify.

Yes. Voluntary donations via the donate page support hosting. We don't sell results, don't sell ads aimed at your inbox content (we can't see it), and don't charge a subscription to find subscriptions.

Optional Google AdSense slots help cover hosting. They only load if you accept the cookie banner — declining means no AdSense script, no advertising cookies, no ads. The scan itself is identical either way.

The parser, MIME decoder, and classifier are versioned in this repo. Licenses for all major dependencies are listed on the licenses page. If you spot a bug or want to contribute a detection rule, check the repo first — most pattern fixes are 5-line edits.