Extract Bank Statements & Invoices at Zero-Token Speed.
Parse complex checking statements, multi-page PDFs, and crinkled receipts into mathematically reconciled, structured JSON & CSV. Sub-50ms deterministic AST parsing for known banking formats with zero token cost, plus neural multimodal AI fallback.
Experience Real-Time Document Extraction
Select one of the pre-loaded benchmark statements below or drag and drop your own PDF or receipt to test the hybrid extraction pipeline in real time.
| Date | Flow | Amount | Description | Category | Account |
|---|
// Extracted JSON Schema will render here
// CSV representation will render here
The 2-Tier Extraction Architecture
How NexiaTK eliminates 90% of recurring AI token bills while guaranteeing 100% mathematical reconciliation on banking statements.
Secure Ingestion Gateway
Requests pass through a zero-trust ingestion gateway enforcing strict cryptographic MIME validation, intelligent token-bucket rate limiting, and ephemeral memory streaming with zero persistent disk retention.
- Zero persistent file storage (RAM buffer only)
- High-throughput memory buffer streaming
- Strict rate limiting & anti-abuse checks
Tier 1: Deterministic Coordinate AST
Known statement layouts (BCA, Mandiri, Wise, Revolut, Jago, Superbank, BNI, CIMB, Danamon) and structured interbank files (SWIFT MT940, OFX, QFX) are parsed natively in milliseconds with 0 AI tokens and 100% mathematical precision.
- 0 AI Tokens used (Completely free compute)
- Sub-40ms deterministic execution latency
- Mathematical balance & checksum validation
Tier 2: Multimodal Neural Vision
When a document is unformatted, handwritten, or photographed from a thermal POS receipt, the system dynamically routes to Gemini 2.5/3.0 Flash for structured visual line-item and tax breakdown extraction.
- Handles crumpled thermal receipts & invoices
- Sub-second neural vision inference
- Line-item categorization & currency normalization
PII Sanitizer & Dispatcher
Extracted data is sanitized in memory. Sensitive account numbers and private identifying metadata are automatically masked before dispatching to clean JSON, CSV, or synchronizing with internal accounting systems.
- Automated PII account masking
- Immediate CSV / JSON export serialization
- Firefly III & ERP webhook sync ready
Engineered to Outperform Generic Solutions
Comparing NexiaTK DocuExtract against standard cloud document extraction APIs and generic LLM wrappers.
| Feature / Metric | NexiaTK DocuExtract | AWS Textract | Mindee OCR | Generic LLM Wrappers |
|---|---|---|---|---|
| Parsing Latency | < 35ms (AST) / 420ms (AI) | 3,500ms โ 6,000ms | 2,000ms โ 4,500ms | 4,000ms โ 10,000ms |
| Cost per 1,000 Statements | $0.00 (Known Banks) | $15.00 โ $50.00 | $20.00 โ $40.00 | $15.00 โ $30.00 |
| Mathematical Balance Check | โ Native Balance Reconcile | โ Raw text only | ~ Partial heuristic | โ Prone to hallucination |
| Universal Bank Standards (SWIFT MT940, ISO 20022 CAMT & OFX) | โ Native Sub-5ms Ingestion | โ Not supported | ~ Requires manual add-on | โ Not supported |
| Southeast Asian & Global Bank Coverage | โ DBS, OCBC, Maybank, CommBank, Chase, BofA, Wise, Revolut, BCA, Mandiri, BNI, CIMB, Jago, Superbank, CAMT.053 | โ English/US biased | ~ Generic Latin only | ~ Inconsistent parsing |
| Data Privacy & Retention | Zero-Log (Memory Only) | Stored in AWS S3 / Cloud logs | Cloud server processed | Third-party model logging |
Integrate with 3 Lines of Code
Integrate DocuExtract directly into your billing platforms, accounting bots, or internal backend services using our clean REST API.
curl -X POST "https://nexiatk.com/api/extract.php" \
-H "X-API-Key: your_live_api_key" \
-F "file=@statement_july_2026.pdf"
import requests
url = "https://nexiatk.com/api/extract.php"
headers = {"X-API-Key": "your_live_api_key"}
with open("statement_july_2026.pdf", "rb") as f:
response = requests.post(url, headers=headers, files={"file": f})
data = response.json()
print(f"Extracted {data['data']['count']} transactions in {data['data']['telemetry']['latency_ms']}ms")
for tx in data['data']['transactions']:
print(f"[{tx['date']}] {tx['type'].upper()} Rp {tx['amount']:,.2f} - {tx['description']}")
<?php
declare(strict_types=1);
$ch = curl_init('https://nexiatk.com/api/extract.php');
curl_setopt_array($ch, [
CURLOPT_POST => true,
CURLOPT_HTTPHEADER => ['X-API-Key: your_live_api_key'],
CURLOPT_POSTFIELDS => [
'file' => new CURLFile(__DIR__ . '/statement_july_2026.pdf', 'application/pdf')
],
CURLOPT_RETURNTRANSFER => true,
]);
$response = curl_exec($ch);
curl_close($ch);
$result = json_decode((string) $response, true);
echo "Extracted: " . $result['data']['bank_name'] . " (" . $result['data']['count'] . " items)\n";
import fs from 'node:fs';
const formData = new FormData();
const fileBlob = new Blob([fs.readFileSync('statement_july_2026.pdf')]);
formData.append('file', fileBlob, 'statement_july_2026.pdf');
const res = await fetch('https://nexiatk.com/api/extract.php', {
method: 'POST',
headers: { 'X-API-Key': 'your_live_api_key' },
body: formData
});
const data = await res.json();
console.log(`Parsed ${data.data.count} items in ${data.data.telemetry.latency_ms}ms`);
Predictable Tiers for Every Scale
Start free in our community sandbox, scale with dedicated API keys, or deploy completely self-hosted on your private infrastructure.
Ideal for evaluating our live engine and parsing occasional personal statements.
- โ 15 Extractions / hour
- โ All deterministic banking parsers
- โ Interactive Web Playground
- โ Direct CSV & JSON download
- โ Memory-only zero log privacy
For fintech apps, accounting tools, and automated backend ingestion workflows.
- โ 10,000 Documents / month
- โ Dedicated API Key & Sub-50ms SLA
- โ Unlimited deterministic bank parsing
- โ Multimodal AI vision fallback included
- โ Automatic webhooks & balance check
- โ Priority email & Telegram support
Complete on-premise Docker deployment for institutions requiring total air-gapped data sovereignty.
- โ Unlimited Extractions
- โ Full Docker Compose & source package
- โ 100% on-premise data sovereignty
- โ Custom bank parser coordinate modules
- โ Direct Firefly III & ERP webhook sync
- โ 24/7 dedicated engineering support