Every record, parsed in 30 seconds.
Drop a certificate — a birth record, a marriage certificate, an official form — and get back structured JSON with every field, confidence scores, and its reference numbers. No setup required.
One call in, one record out.
A clean JSON envelope with every field, a per-field confidence score, the document's reference and registration numbers, and a request id you can quote in support tickets.
- Document type
- Reference number
- Registration number
- Full name
- Date of birth
- Place of issue
- Date issued
- Issuing authority
- Certification status
{
"success": true,
"request_id": "req_7d2f9a…",
"latency_ms": 1510,
"fields": {
"document_type": { "value": "BIRTH CERTIFICATE", "confidence": 99.8, "band": "verified" },
"reference_number": { "value": "BC-2026-0098471", "confidence": 99.2, "band": "verified" },
"full_name": { "value": "ANGELA ZOE", "confidence": 99.0, "band": "verified" },
"date_of_birth": { "value": "1995-01-01", "confidence": 99.7, "band": "verified" },
"registration_no": { "value": "REG-774120", "confidence": 98.4, "band": "high" },
"issuing_authority": { "value": "GENERAL REGISTER OFFICE", "confidence": 99.5, "band": "verified" }
}
}The documents most OCRs quietly skip.
Certificates and civil records are where template-based OCR falls apart — ornate borders, stamps, handwriting, and a different layout in every jurisdiction. OnPoint reads them the same way it reads a passport: end to end, no templates.
Any certificate, any layout
Birth, marriage, and death certificates, civil records, and official forms — read end to end with no per-issuer template. Region-aware field logic adapts to the document in front of it.
Reference & registration numbers
The identifiers your systems key on — reference numbers, registration numbers, certificate serials — captured exactly as printed, with their own confidence score.
Sub-second median latency
P50 ≈ 1.4 s. P99 below 3.2 s on production traffic. No queueing, no warmup — built for synchronous onboarding and case work.
Mapped to your schema
Every record comes back in the exact schema your systems already expect — set your field names, casing, and date formats once, and every document family returns in that shape.
Per-field confidence scores
Five confidence bands per field, not per document. Auto-route the one uncertain field to a reviewer instead of failing the whole call.
Zero-retention by default
Images and extracted values are processed in-memory and discarded the moment your response is returned. Enforced at the application layer.
If it's an official record, we read it.
Certificates, civil records, and government forms from any issuing authority — benchmarked continuously against fresh in-the-wild documents. One integration, mapped to the schema your systems already use.
One API key. One balance. Every document family.
Government OCR runs on the same endpoint, the same authentication, and the same credit balance as Passport and ID OCR. Send a certificate today with the key you already have — nothing new to provision.
Free for your first 5 documents. No card.
Test the full production API end-to-end before you ever talk to billing. It's the same credit that runs passports and IDs — one balance, every service. Volume pricing kicks in automatically when you keep going.
Put it in front of real government documents.
One API key, every certificate and civil record, in production this afternoon.