Match invoices to bank transactions
in seconds.

PDF, CSV, XLSX, XLS, OFX, QFX, MT940, and BAI2. For Excel files, we use SheetJS to convert to text before AI extraction. For PDF statements, the AI reads directly from the document.

Yes. Upload statements from any number of banks. Each statement is processed independently and transactions are matched against your full invoice database.

Stage 1 is a fast algorithmic filter: it checks currency match, date proximity (-7 to +60 days), and amount tolerance (±10%). Stage 2 sends candidate pairs to OpenAI GPT, which evaluates vendor names, descriptions, and context to confirm or reject each match.

Unmatched transactions are listed in a dedicated "Unmatched" tab. You can create manual matches by selecting a transaction and choosing an invoice from your database.

No. Reconciliation matching is included free with every bank statement upload. Only the bank statement extraction itself costs 1 credit.

Yes. The AI detects the currency of each transaction and can match transactions in one currency with invoices in another, accounting for exchange rate differences.

The confidence score (0.0 to 1.0) indicates how certain the AI is about a match. Scores above 0.6 are presented as suggestions. Each match also includes an AI-generated explanation of why the match was proposed.

Currently, all matches require manual confirmation. This ensures accuracy and gives you full control. Confirmed matches automatically set the corresponding invoice status to "paid".

Most bank statements are processed within 1-2 minutes regardless of size. The matching pipeline runs automatically after extraction — by the time you open the matches tab, results are usually ready.

Yes. Use the "Run Reconciliation" button to re-process unmatched transactions against your current invoice database. This is useful when invoices arrive after the bank statement was uploaded.