Turn raw UPI strings into
structured intelligence
One API call turns messy bank transaction descriptions into clean merchant names, categories, and spending signals — built specifically for Indian fintech.
Every team building on India's payment stack
Whether you're underwriting credit, building expense views, or integrating UPI data into your product — UPIForge is the intelligence layer your stack is missing.
Credit Underwriters
Raw UPI strings tell you nothing. Categorized transaction data tells you how a borrower actually spends — groceries vs. EMIs vs. impulse shopping.
Expense Management Products
Your users see "UPI/DR/98234/ANANDA BHA/paytmqr@" in their statement. You need to show them "Food & Dining — ₹450". That gap is what we close.
Personal Finance & Wealth Apps
Spending insights, category budgets, and cash flow summaries all require the same foundation — structured transaction data. Start with that, not from scratch.
Teams Who Want to Integrate Fast
One REST endpoint. JSON in, JSON out. No model training, no regex maintenance, no edge cases to handle. Drop in the API key and ship.
Everything your app needs from a transaction
A multi-stage intelligence pipeline — rule engine, ML classifier, and LLM fallback — that handles the full chaos of Indian UPI and bank strings.
Merchant Normalization
Maps truncated, abbreviated UPI strings to canonical merchant names. Handles IndusInd truncation, Paytm QR codes, NEFT references, and regional Indian merchants.
Category & Subcategory
15 L1 categories, 50+ L2 subcategories. Powered by a merchant registry, ML classifier, and LLM fallback for unknown merchants.
Recurring Detection
Automatically identifies EMIs, subscriptions, rent, and utility bill cycles. Essential for cash flow views and credit underwriting.
Income Classification
Detects salary, freelance income, interest, and refunds from CREDIT transactions. Distinguishes income signals from P2P transfers.
Confidence Scoring
Every result carries a confidence score. Registry hits score 0.90+, ML hits 0.70+, LLM fallback 0.55–0.65. Build rule-based UX on top.
Batch Processing
Process up to 5,000 transactions in a single API call. Results return in milliseconds per transaction — production-grade throughput.
Simple REST API
Authenticate with an X-API-Key header.
All endpoints return JSON. Contact us to get an API key.
Enrich a single transaction. Returns merchant, category, subcategory, confidence, recurring flag, and income flag.
Enrich up to 5,000 transactions in one call. Ideal for statement processing, demo reports, and bulk enrichment.
Service health check. Returns version, uptime, and model load status. No auth required.
Readiness probe. Returns 200 only when database and ML models are fully loaded.
Start with a free evaluation
Send us a sample of 500 anonymised transactions. We'll run them through the API and send you a full intelligence report — category breakdown, merchant coverage, confidence distribution — in 48 hours. No strings attached.
Request Free EvaluationWe respond within 24 hours. Evaluation is free, no commitment required.