Convert to E-Invoice

Upload your PDF, CSV, or Excel invoice. Our AI extracts the data and converts it to Peppol, XRechnung, or Factur-X format.

Drop your invoice here

or click to browse

PDF CSV / Excel Image
Or paste invoice text

Secure Cloud Processing

Encrypted in transit • EU servers • Auto-deleted after processing • No human review

Learn more

AI-Powered

Claude extracts all fields

Review & Edit

Verify before generating

Multi-Format

Peppol, XRechnung, Factur-X

Conversions count as 2 validations from your quota

How it works

Step 1

Upload

PDF, image, CSV, or Excel. Up to 10 MB. Printed or scanned PDFs both work.

Step 2

Extract + Review

AI reads invoice data, you confirm in an editable form. Uncertain fields flagged.

Step 3

Generate + Validate

Pick target format. Engine generates XML, validates against EN 16931 + Schematron, auto-fixes structural issues.

Step 4

Download

On compliance success: XML + diff + Evidence Pack. €7.90 PAYG per file.

Frequently Asked Questions

Which file types can I upload?

PDF (printed or scanned), CSV, Excel (.xlsx, .xls), and image files (PNG, JPEG). For PDF and image input, our engine uses Anthropic Claude Vision to extract structured invoice data. For CSV/Excel, the column mapping happens via UI on first upload and is remembered for subsequent uploads. Max 10 MB per file.

Which output formats do you support?

XRechnung 3.0.2 (UBL or CII), ZUGFeRD 2.3.2 (Minimum / Basic / EN 16931 / Extended profiles, hybrid PDF/A-3 output), Factur-X 1.07.2 (same spec as ZUGFeRD, French branding), Peppol BIS Billing 3.0, and generic EN 16931 UBL or CII. The engine validates output against the appropriate Schematron before download.

Is conversion free?

Yes. Upload, extract, review, generate, validate — all free. Unlimited, no account, no caps. You only pay to download the final repaired XML + diff report + Evidence Pack: €7.90 per file (PAYG; €29 for 5, €79 for 20). Pay only on success — if our engine cannot turn your input into a compliant XML, nothing is charged. See /pricing for details.

How accurate is the AI extraction?

We use Anthropic Claude Vision for PDF/image extraction. Accuracy varies with source quality — typical print-based PDFs reach >90% field accuracy; scanned PDFs at 300+ DPI are ~85%. Critical: you always see the extracted values in an editable review form before XML is generated. Uncertain values are flagged. We don't silently assume — you confirm.

Where does my data go? Is the PDF stored?

PDF/image data is sent to Anthropic Claude Vision (US, EU-US Data Privacy Framework compliant). The raw file is NOT retained — only the extracted structured data, for your review. Anthropic does not use your data to train models (per their API ToS). If you download an Evidence Pack, the extracted data is retained for 12 months for re-download (anonymized after 30 days). Full policy: /privacy.

What if I have just an XML file (no PDF)?

Use the validator instead: /validator. It accepts existing XML, validates against EN 16931 + format-specific Schematron, auto-fixes structural errors, and produces the same Evidence Pack on download. The two flows share the engine — same canonical model, same validation, same fix logic.

Related