Word/line bounding boxes
Every recognized text location keeps its position. That yields accountable extraction for review, DATEV and agents — not plain text without geometry.
Made for real business. Not for hype.
API output from July 14, 2026
Source document (invoice) · from €9 per 1,000 pages
Extracted fields · from €9 per 1,000 pages
| # | SKU | Description | Unit | Qty | Rate | Amount |
|---|---|---|---|---|---|---|
| 1 | ITEM-92912 | Multifold paper towels (case, 16 packs) | case | 40 | $42.11 | $1,684.40 |
| 2 | PN-61493 | Foaming hand soap refill, 1200 mL (case of 4) | case | 4 | $38.81 | $155.24 |
| 3 | PART-33106 | 55-gallon trash can liners (case of 100) | case | 9 | $53.07 | $477.63 |
LLM fields — natural language
Accounts payable
Prompt
Based on the total amount, check whether this invoice exceeds the dual-control limit of USD 2,000 and therefore needs a second approval.
Result
Yes — USD 2,527.98 exceeds USD 2,000. Second approval required.
Document INV-59762 · field _total_amount
KYC / HR / legal capacity
Prompt
Based on the date of birth on the ID document, check whether the person is of legal age (18) as of today.
Result
Yes — date of birth 12 Mar 2004 → of legal age on the reference date (22 years).
ID / passport pipeline · LLM field instead of a fixed template key
Compliance / vendor screening
Prompt
Based on supplier name and address, check whether the issuer is located in the USA and must be treated as a third-country vendor for EU booking.
Result
Yes — Pristine Galvan Inc., Cleveland OH (USA). Third country / US vendor.
Document INV-59762 · address + company name via LLM prompt
Agent UI: extracted data (raw JSON)
JSON excerpt (7 fields + 3 line items)
{
"_supplier_name": {
"value": "Pristine Galvan Inc.",
"value_raw": "Pristine Galvan Inc."
},
"_invoice_number": {
"value": "INV-59762",
"value_raw": "INV-59762"
},
"_invoice_date": {
"value": "07/07/2026",
"value_raw": "2026-07-07"
},
"_currency": {
"value": "USD",
"value_raw": "USD"
},
"_net_amount": {
"value": "$2,317.27",
"value_raw": "2317.27"
},
"_vat_amount": {
"value": "$162.21",
"value_raw": "162.21"
},
"_total_amount": {
"value": "$2,527.98",
"value_raw": "2527.98"
},
"_line_items": [
{
"sku": "ITEM-92912",
"description": "Multifold paper towels (case, 16 packs)",
"qty": 40,
"unit": "case",
"rate": "$42.11",
"amount": "$1,684.40",
"amount_raw": "1684.40"
},
{
"sku": "PN-61493",
"description": "Foaming hand soap refill, 1200 mL (case of 4)",
"qty": 4,
"unit": "case",
"rate": "$38.81",
"amount": "$155.24",
"amount_raw": "155.24"
},
{
"sku": "PART-33106",
"description": "55-gallon trash can liners (case of 100)",
"qty": 9,
"unit": "case",
"rate": "$53.07",
"amount": "$477.63",
"amount_raw": "477.63"
}
]
} In the same account
AI OCR with bounding boxes — 69 types, fields with location.
from €9 per 1,000 pages
Seamless in the job path — task and document also on mobile.
Web + Mobile
SES signatures from the same credit balance.
SES from the balance
Audit-proof filing and audit trail.
included
Our prices are public — including partner terms. Transparent list price for partners and integrators.
Unit: Price per 1,000 pages
| Tier | Elite Partner (−70%) | Partner (−40%) | Enduser |
|---|---|---|---|
| Basic | €9.00 | €18.00 | €30.00 |
| Premium | €12.00 | €24.00 | €40.00 |
| Ultra | €30.00 | €60.00 | €100.00 |
Billing is internal in credits; amounts follow the selected currency (base: list price).
Elite terms after qualification — criteria and program are public: Pricing · Partner program
Trusted by leading companies worldwide
Every recognized text location keeps its position. That yields accountable extraction for review, DATEV and agents — not plain text without geometry.
Geometry becomes a searchable archive PDF — ideal for audit-proof filing and later full-text search.
Ambiguous cases can be clicked and corrected on the document. The correction stays on the object — including an audit trail.
The same extraction is reachable via REST and the 300+ API/MCP Tools — one account, one credit balance.
AI OCR recognizes text in documents with geometry (bounding boxes) and can produce searchable archive PDFs — beyond classic text recognition.
Bounding boxes are coordinates per word or line. They keep extraction accountable to a location on the page.
A searchable PDF where recognized text sits as an invisible layer over the scan — ideal for archives and full-text search.
The feature page explains the API capability; the product page covers the full OCR/IDP offering. The feature URL canonicalizes to the product.
Yes. The same extraction is reachable via REST and MCP.
Over 120 OCR languages — depending on pipeline and document quality.
Typically per page or credits; Elite pipelines from about €9 per 1,000 pages.
In PaperOffice EU infrastructure on owned hardware.
Yes — human-in-the-loop allows visual review and correction on the document.
For teams and integrators that need text recognition with geometry, archive PDF, and API/MCP.