In development · Early access

Catch invoice errors before they cost you.

Dinasty Invoice reviews supplier invoices before payment. It flags likely duplicates, missing data and inconsistencies, then explains each alert in plain language with Claude. Your team makes every decision.

Request early access How it works
Review · 5 invoicesDemo preview · fictitious data
High · Possible duplicate

Same supplier and same invoice number: INV-1042 and INV-01042.

Human check: compare both PDFs before paying either.
Medium · Possible duplicate

Same supplier and amount (389.50 EUR), 4 days apart, different numbers.

Human check: confirm with the supplier.
Medium · Missing data

Invoice number and date not found.

Human check: ask the supplier for a corrected invoice.
Features

What Dinasty Invoice checks

We are building in the open. Each feature shows its real status today.

Prototype

Duplicate detection

Same supplier and number, even with different formatting. Same amount within a few days.

Prototype

Missing & inconsistent data

Missing supplier, number, date or total. Zero amounts and future dates.

Prototype

Plain-language explanations

Claude explains why each alert matters and what to check. It never invents amounts.

Prototype

PDF extraction

Supplier, number, date, amount and currency from text-based PDFs.

Planned

Scanned invoices

OCR for image-only PDFs and photos.

Planned

Purchase-order matching

Compare invoices with orders and VAT rules.

How it works

From PDF to a reviewed report

  1. Import

    Drop in one or more PDF invoices.

  2. Extract

    Key fields are read and normalized. Unclear fields stay "Unknown".

  3. Check

    Deterministic rules flag duplicates and gaps, with the evidence.

  4. Review

    Claude explains each alert. A person decides what happens next.

Rules find, AI explainsAlerts come from reproducible rules. Claude does not create them.
Human approval, alwaysDinasty never pays, sends reminders or edits accounting data.
Minimal dataOnly extracted fields are sent to the AI, after explicit consent.
No retention by defaultDocuments are processed in memory and not stored.

Join the early access

We are looking for finance and operations teams who want to test the prototype and shape the product.

Request early access pro.jflan@dinasty.tech