A chain you can actually read
Two providers, both EU-incorporated, under Article 28 DPAs — and we publish their sub-processors too, which is where most EU-only claims quietly fall apart.
In development · processed in the EU
Structured data out of documents
Invoices, contracts and forms become typed JSON with a confidence score on every field. Built so a compliance review is a short conversation instead of a six-week blocker.
No API keys yet · no signup · nothing to cancel
POST /v1/parse
Illustrative response · the API is not open yet
{
"supplier": { "value": "Acme B.V.", "confidence": 0.99 },
"vat_id": { "value": "NL812345678B01", "confidence": 0.98,
"checksum_valid": true },
"invoice_date": { "value": "2026-08-14", "confidence": 0.94 },
"iban": { "value": "NL91ABNA0417164300", "confidence": 0.86 },
"total_gross": { "value": 1284.50, "currency": "EUR" },
"checks": { "line_items_sum_matches_total": true }
}
{
"counterparty": { "value": "Meridian GmbH", "confidence": 0.99 },
"effective": { "value": "2026-09-01", "confidence": 0.98 },
"term_months": { "value": 24, "confidence": 0.92 },
"notice_days": { "value": 90, "confidence": 0.88 },
"auto_renew": { "value": true, "confidence": 0.97 },
"checks": { "notice_before_renewal": true }
}
≥97% certain
90–96% check
<90% review
Invoices
Contracts
Forms
Your own schema
0
third-party requests on this page
0
cookies set
2
providers in the processing chain
EU
where every document is processed
Why another document API
Built for people who answer to a compliance team
Your schema, typed
Define the fields and types you want back. Get those, or an explicit typed error — not a wall of raw text to post-process.
Confidence on every field
Each value carries its own score. Low-certainty fields are marked in the response so review is a glance, not a re-read of the source document.
Validation, not just extraction
Line items sum to the total. VAT IDs checksum. Dates make sense. A document whose numbers do not reconcile is flagged, not returned as fact.
Deleted by default
Documents are deleted once parsing finishes. Zero-retention mode for anything that cannot touch disk at all.
Cheap pages stay cheap
Most pages never reach a paid model. Escalation happens per page rather than per document, and the reason is in the response.
Compliance
The boring version, in full
Where documents are processed, who can be compelled to hand them over, and which claims do not survive scrutiny. Written down before launch rather than after the first security questionnaire.
Processing locationGermany (EU)
Contracting entitiesEU-incorporated
Providers in the chain2, both published
Their sub-processorsPublished too
Training on your documentsNever
CertificationsNone yet
Breach notice72 h · GDPR Art. 33
Pricing
Indicative
Pay for pages, not seats
Not final. It is here so you can tell me whether it is plausible. Design partners from this page get launch pricing held.
| Plan | Monthly | Included |
|---|---|---|
| Free | € 0 | Evaluation volume, no card |
| Starter | € 29 | Low volume, single schema |
| Growth | € 99 | Batch endpoint, webhooks, multiple schemas |
| Scale | € 299 | Higher throughput, priority support |
Status
Built
Being decided
Not claimed
What is actually true today
Nothing you can call yet. This is deliberate — the page exists to find the people who need it before the API is fixed in place.
Where each provider physically runs inference, and their retention terms. Verified before anything ships, not after.
No certifications, no signed DPA, no uptime history. Those get built and published — they are not being implied here.
If a data-residency or compliance requirement has ever stopped you from using a document processing API, I would like to hear how that actually went — even if you would never buy this.
Try it on your worst document
Leave an email and I will follow up personally. No sequence, no newsletter, no sales call.