Questions people actually ask
Straight answers, including where the honest limits are.
What kinds of bank statements can I upload?
Digital PDF exports from online banking, scanned or photographed PDFs, and direct photos (JPEG/PNG/WebP) — a phone photo of a paper statement works too. Each page is read as text where possible, and falls back to vision-based reading automatically when there's no text layer to work from.
What about cash purchases that never show up on a bank statement?
Photograph the paper receipt and add it as a cash receipt — Claude reads the merchant, date, and total, and it's categorized and included in your reports exactly like any other transaction, tagged with a "Receipt" badge so you can tell it apart from a bank statement.
Does it work with statements from any bank?
Extraction reads the actual content of each page rather than relying on a fixed template for one specific bank, so it adapts to different layouts. That said, statement formats vary a lot — always worth spot-checking the first upload from a new bank against the original.
How accurate is the extraction?
Every statement is cross-checked automatically: the totals we extract from individual transactions are compared against the bank's own printed summary figures, and any mismatch is flagged clearly on the Analyze page. That catches missed or misread transactions rather than leaving them silent — but for anything you're relying on for tax or compliance purposes, we'd still recommend a human spot-check.
Can I fix a miscategorized transaction?
Yes — change it directly from a dropdown in the Analyze table. That correction is remembered, retroactively applied to every other transaction with the exact same description, and used as guidance the next time you upload a statement, so the same merchant gets recognized correctly going forward.
Can I fix several transactions at once, not just one at a time?
Yes — tick the checkboxes on any set of transactions in the Analyze table, even ones that don't share a description, and move them all to a new category in one action. The same idea exists for payees too: select several groups on the Payees page and merge them into one in a single click, rather than merging pairs one at a time.
Why do some payments to the same business show up as different payees?
A date or reference number embedded in the description (e.g. "On 09 Jun" vs "On 19 Jun") is stripped automatically, so those group together on their own. But genuinely different-looking descriptions for the same real business — their card payments vs. their website payments, say — need a one-time manual merge: type a snippet of the other description and it's remembered for good, and several groups can be merged together in one action.
Can I leave a note on a specific transaction?
Yes — any transaction can carry a short note, always visible under its description, carried through to every export format. Unlike category corrections, a note stays specific to that one line rather than applying to every transaction with the same description.
Can it help with VAT?
Yes — each category can carry a VAT rate (standard, reduced, zero-rated), and the Analyze page shows a live output/input/net VAT estimate built from that. Clearly labeled as an estimate, not a filed return — real VAT treatment depends on the actual invoice, not just which category a transaction landed in, so anything genuinely ambiguous is left unset for manual review rather than guessed.
Can I create my own categories?
Yes, fully — add, rename, recolor, or delete categories per client. Renaming or deleting updates your existing transactions automatically, so nothing's left pointing at a category that no longer exists.
I manage statements for more than one business — is that supported?
Yes. Each client is a fully separate workspace — its own statements, categories, and totals — switchable from a picker in the sidebar. A correction or category change on one client's books never affects another's.
I'm an accountant or bookkeeper — is this actually built for that, not just personal use?
Yes, deliberately. Challenger banks (ANNA, Monzo, Countingup) already offer something like this natively, but only for clients who bank with them directly — real practice work means clients spread across many different banks. Extraction reads any statement's actual content rather than a template built for one bank, so it isn't limited to one provider's own customers.
How do I know I can trust what the AI decided, without checking everything by hand?
You don't have to take it on faith — every categorization is a visible, editable dropdown right in the ledger, and every statement's extracted totals are automatically checked against the bank's own printed summary, with any mismatch flagged clearly. Fix something once and it's remembered going forward, so the more you use it, the less you need to double-check.
What can I export, and what does it include?
An interactive HTML report you can click through offline, a branded PDF (cash flow, categories, flagged items, account details, full ledger, plus payees and payers), a formatted Excel workbook, or CSV/JSON — whichever suits sending to a client, an accountant, or feeding into another tool.
Is my data secure?
Passwords and API keys are hashed, every session is scoped strictly to your own account and client data, and the app runs over HTTPS with standard security headers (CSP, HSTS, and others). Account details (account number, IBAN, sort code, holder name/address) and the original uploaded file itself are both encrypted at rest with AES-256-GCM — a database compromise doesn't hand over readable statement data. Being self-hosted, your data stays on infrastructure you control, not a third party's shared servers. For context: this is a self-hosted setup rather than an ISO 27001-certified data centre with independent audits — the right fit for a trusted single-operator or small-team setup, not a substitute for formal compliance certification at enterprise scale.
Is this GDPR compliant?
The technical foundations are genuinely in place: encryption at rest for sensitive fields and original files, self-hosted infrastructure, hashed credentials, a working data export tool, account deletion that removes everything, and a configurable retention period. GDPR compliance itself is ultimately a legal state as much as a technical one, though — it depends on having a proper privacy policy, a documented legal basis for processing, and (if you're a bookkeeper processing clients' data) a Data Processing Agreement in place. Draft templates for all three are linked in the footer, ready for a solicitor's review — responsibility for compliance sits with whoever operates the app, not the software itself.
Question not covered here? Get in touch — we'd rather give you a straight answer than have you guess.