Mexican Document OCR API: Extract Data from INE, CFDI, CSF, and CURP

Mexico's regulatory landscape requires businesses to process several types of government-issued documents: voter IDs (INE), digital invoices (CFDI), tax certificates (CSF), and population registry documents (CURP). Manually extracting data from these documents is slow and error-prone. This guide covers how to automate extraction for all of them with a single API.

Supported Mexican documents

Extract INE voter ID data

Python

from ocilar import OcilarClient

client = OcilarClient(api_key="sk-your_key")

result = client.extract_ine(
    front="ine_front.jpg",
    back="ine_back.jpg"
)

print(result.nombre)           # "JULIO JUAREZ MARTINEZ"
print(result.curp)             # "JUAJ850101HMCRRL09"
print(result.fecha_nacimiento) # "1985-01-01"
print(result.domicilio)        # "CALLE INDEPENDENCIA 123..."
print(result.vigencia)         # "2029"

# Check if expired
from datetime import datetime
is_valid = int(result.vigencia) >= datetime.now().year

cURL

curl -X POST https://api.ocilar.com/api/v1/extract/ine \
  -H "X-API-Key: sk-your_key" \
  -F "front=@ine_front.jpg" \
  -F "back=@ine_back.jpg"

Extract CFDI invoice data

Python

result = client.extract_cfdi(file_path="factura.xml")

print(result.uuid)          # "6128a3d4-1234-..."
print(result.emisor_rfc)    # "AAA010101AAA"
print(result.total)         # 11600.00
print(result.iva)           # 1600.00

for concepto in result.conceptos:
    print(concepto.descripcion, concepto.importe)

From bytes (e.g. downloaded from SAT portal)

with open("factura.xml", "rb") as f:
    result = client.extract_cfdi(file_bytes=f.read())

# Supports CFDI 3.3 and 4.0 automatically

Extract CSF (Constancia de Situación Fiscal)

Python

result = client.extract_csf(file_path="csf.pdf")

print(result.rfc)                  # "JUAJ850101ABC"
print(result.nombre)               # "JULIO JUAREZ MARTINEZ"
print(result.regimen_fiscal)       # "612 - Personas Físicas con Actividades Empresariales"
print(result.domicilio_fiscal)     # "CALLE INDEPENDENCIA 123..."
print(result.codigo_postal)        # "44100"
print(result.actividades)          # ["Desarrollo de software", ...]
print(result.fecha_inicio_ops)     # "2018-03-01"

Node.js

import { OcilarClient } from '@ocilar/sdk'
import { readFileSync } from 'fs'

const client = new OcilarClient({ apiKey: 'sk-your_key' })

const result = await client.extractCsf({
  file: readFileSync('csf.pdf')
})

console.log(result.rfc)
console.log(result.regimenFiscal)
console.log(result.domicilioFiscal)

Extract CURP card

from ocilar import OcilarClient

client = OcilarClient(api_key="sk-your_key")

result = client.extract_curp(file_path="curp.pdf")

print(result.curp)             # "JUAJ850101HMCRRL09"
print(result.nombre_completo)  # "JULIO JUAREZ MARTINEZ"
print(result.fecha_nacimiento) # "1985-01-01"
print(result.entidad)          # "JALISCO"
print(result.sexo)             # "H"

Multi-document onboarding flow

A complete KYC onboarding flow for a Mexican fintech typically requires INE + CSF (for business customers) or INE + CURP (for individuals). Here's a complete example:

from ocilar import OcilarClient
from datetime import datetime

client = OcilarClient(api_key="sk-your_key")

def kyc_onboard_individual(ine_front: str, ine_back: str, curp_pdf: str) -> dict:
    """Full KYC extraction for individual customers."""

    # Extract INE
    ine = client.extract_ine(front=ine_front, back=ine_back)

    # Validate INE not expired
    if int(ine.vigencia) < datetime.now().year:
        raise ValueError(f"INE expired in {ine.vigencia}")

    # Extract CURP to cross-validate
    curp = client.extract_curp(file_path=curp_pdf)

    # Cross-check: CURP on INE should match CURP document
    if ine.curp != curp.curp:
        raise ValueError("CURP mismatch between INE and CURP document")

    return {
        "nombre": ine.nombre,
        "curp": ine.curp,
        "fecha_nacimiento": ine.fecha_nacimiento,
        "domicilio": ine.domicilio,
        "estado": ine.estado,
        "ine_vigencia": ine.vigencia,
        "verified": True
    }

def kyc_onboard_business(ine_front: str, ine_back: str, csf_pdf: str) -> dict:
    """Full KYC extraction for business customers (persona moral/fisica con actividad empresarial)."""

    rep_legal = client.extract_ine(front=ine_front, back=ine_back)
    empresa = client.extract_csf(file_path=csf_pdf)

    return {
        "representante_legal": rep_legal.nombre,
        "rfc_empresa": empresa.rfc,
        "razon_social": empresa.nombre,
        "regimen_fiscal": empresa.regimen_fiscal,
        "domicilio_fiscal": empresa.domicilio_fiscal,
        "actividades": empresa.actividades,
        "verified": True
    }

Industry use cases

Fintech & lending (KYC)

Mexican regulation requires identity verification for any credit product. Extract INE data automatically during the loan application flow — no manual form filling for the borrower, no manual review for your team.

B2B vendor onboarding

Extract and validate supplier RFC and fiscal regime from their CSF before adding them to your accounts payable system. Ensure you're paying the correct RFC and catching mismatches early.

Accounting software & ERPs

Parse CFDIs received from vendors automatically. Extract UUID, amounts, line items, and fiscal data directly into your accounting module without copy-pasting from PDFs.

HR & payroll platforms

Onboard new employees by scanning their INE and CURP. Feed data directly into IMSS and INFONAVIT systems without manual data entry.

Sharing economy & gig platforms

Verify driver, delivery agent, and contractor identities with INE extraction as part of onboarding. Pair with IMSS verification to confirm employment history.

Pricing

FAQ

What image quality is required for INE?

Minimum 300 DPI equivalent. The card must be fully visible, well-lit, no blur, no glare. Mobile photos work well when the card fills most of the frame.

Does CFDI extraction support CFDI 3.3 and 4.0?

Yes. Both versions are handled automatically — you don't need to specify the version.

Can I extract data from scanned CSF PDFs?

Yes. The CSF extractor handles both digital PDFs (from SAT portal) and scanned copies, though digital PDFs produce higher accuracy.

Is there an SDK for Go or PHP?

Python and Node.js SDKs are available. REST API works with any language. See the full documentation.