Bank-statement intelligence API · Built for India
Tarazu turns PDF and Excel bank statements into clean transactions, explains suspicious activity with a risk score and reasons, checks the document for tampering and extracts verified account details — in one API call.
statement_apr_sep.pdf
A/c XXXXXXXX4417 · Savings
| Date | Narration | Debit | Credit | Balance |
|---|---|---|---|---|
| 03 Sep | UPI/CR/4471/ANKxx@ybl | 9,500 | 10,120 | |
| 03 Sep | UPI/CR/4472/MEHxx@okaxis | 14,900 | 25,020 | |
| 03 Sep | IMPS/DR/BENE-07 | 24,400 | 620 | |
| 04 Sep | SALARY/ACME LOGISTICS | 31,250 | 31,870 | |
| 04 Sep | UPI/CR/5530/RAJxx@paytm | 10,000 | 41,870 | |
| 04 Sep | UPI/CR/5531/SURxx@ibl | 10,000 | 51,870 | |
| 05 Sep | NEFT/DR/BENE-07 | 51,200 | 670 | |
| 05 Sep | ATM/WDL/0941 | 500 | 170 |
Transactions
—
Integrity
—
Risk score
—
§1 How it works
No scraping, no credentials, no screen-sharing. The customer gives you a statement and consent; you give it to us.
Send a PDF (password-protected is fine), XLS, XLSX or CSV along with the consent reference you captured from your customer.
POST /statements/analyze
Transactions are parsed and categorised, the document is checked for edits, and behaviour is scored against suspicious-activity indicators.
≈ 4 s · 6-month PDF
One JSON with transactions, account details, integrity findings and a risk score with weighted, plain-language reasons. Sync or webhook.
statement.completed
§2 The API
Pick the checks you need per call. Every risk score ships with the reasons behind it, so your team can explain a decision to a customer, an auditor or a regulator.
PASS_THROUGH
92% of credits leave the account within 30 minutes
MANY_TO_ONE
214 distinct UPI senders, 3 recurring beneficiaries
LOW_RESIDUAL
Median end-of-day balance below ₹600
Scores are indicators that support human review. Tarazu never labels a person or account as a mule or a fraudster.
Read the API reference{
"id": "stm_01JD4Q7Y3W",
"status": "completed",
"processing_ms": 4180,
"account": {
"holder_name": "R•••• K•••• S••••",
"bank": "Sample Co-operative Bank",
"account_no": "XXXXXXXX4417",
"ifsc": "SMPL0001234",
"type": "savings",
"period": {
"from": "2026-04-01",
"to": "2026-09-30"
}
},
"summary": {
"transactions": 1284,
"credits_inr": 4821330.5,
"debits_inr": 4809112,
"closing_balance_inr": 18630.25
},
"integrity": {
"verdict": "review",
"checks_passed": 11,
"checks_total": 13,
"findings": [
"font_mismatch_page_4",
"running_balance_break_row_612"
]
},
"risk": {
"score": 78,
"band": "high",
"label": "suspicious_activity_indicators",
"reasons": [
{
"code": "PASS_THROUGH",
"weight": 0.31,
"detail": "92% of credits leave the account within 30 minutes"
},
{
"code": "MANY_TO_ONE",
"weight": 0.22,
"detail": "214 distinct UPI senders, 3 recurring beneficiaries"
},
{
"code": "LOW_RESIDUAL",
"weight": 0.14,
"detail": "Median end-of-day balance below ₹600"
}
],
"disclaimer": "Indicators only. Not a determination of fraud or mule activity."
}
}§3 Products
Use them together in one call or separately. Each one returns structured output you can store, audit and rule on.
checks: ["parse"]Digital and password-protected PDFs, XLS/XLSX and CSV exports into a single clean transaction schema: date, narration, mode (UPI, IMPS, NEFT, NACH, ATM, cheque), counterparty, amount, running balance and category.
checks: ["risk"]A 0–100 score and band built from behavioural indicators commonly associated with money-mule and layering patterns, each with a weight and a sentence you can show a reviewer.
checks: ["integrity"]Catch statements that were edited after download: producer metadata, font and glyph consistency, running-balance continuity, date ordering, duplicate references and page structure.
checks: ["account"]Holder name, masked account number, IFSC, branch, account type and statement period, normalised and ready to match against your KYC and penny-drop data.
Webhooks
Signed callbacks for long statements.
Sandbox keys
Fixture statements, no real data.
Idempotent calls
Safe retries with request keys.
Batch re-screening
Re-score accounts on a schedule.
§4 Who it's for
Pull six months of inflows, salary regularity, EMI obligations and bounce history out of any statement, then attach integrity findings to the credit memo so reviewers see why a file was escalated.
< 10 s
typical end-to-end on a 6-month PDF
Figures are illustrative targets for the preview.
§5 Trust
The platform is designed around India's Digital Personal Data Protection Act, 2023: purpose-bound consent, minimal retention and a clear trail of who saw what.
Every call carries a consent reference captured by you from the data principal. No consent, no processing.
Raw files are deleted once processing finishes. Derived reports expire on a schedule you set, or on demand via API.
TLS 1.2+ in transit, AES-256 at rest, per-tenant keys, and statement passwords that are never written to disk.
Processing and storage planned in Indian data centres, with no third-party model training on your customers' data.
Scoped API keys, IP allowlists, role-based dashboard access and signed webhooks.
Every upload, view, export and deletion is logged against a user or key, ready for your grievance and audit teams.
Default data lifecycle
T + 0
Statement received with consent reference
T + seconds
Parsed, checked, scored; raw file deleted
T + 30 days
Derived report auto-expires (configurable)
Any time
Delete-by-API or on data-principal request
Defaults shown for the preview. Final retention terms will be set in the data processing agreement.
§6 Pricing
We're finalising plans with early partners. Tell us your monthly volume and which checks you need, and we'll share a proposal.
§7 FAQ
No. Tarazu reports suspicious-activity indicators, a risk score and the reasons behind it. It is decision support for your team; the decision and any reporting obligations remain with you.
Digital PDFs (including password-protected ones), XLS, XLSX and CSV exports from Indian banks. Scanned image PDFs are on the roadmap and will be flagged as such rather than silently mis-read.
We compare what a genuine statement should look like with what the file actually contains: producer metadata, embedded fonts and glyph widths, running-balance arithmetic, date ordering, duplicate references and page structure. Each finding points to the page or row involved.
A consent reference you capture from your customer for this specific purpose. It travels with the request and is stored in the audit trail, so you can evidence consent under the DPDP Act.
Raw files are deleted immediately after processing by default. Derived reports expire after a configurable window, and you can delete any report at any time through the API or the dashboard.
Yes. Sandbox keys come with fixture statements that exercise every check and every risk band, so you can build your integration end to end before going live.
§8 Get started
Tell us what you're building. We'll set you up with sandbox keys, fixture statements and a call with someone who has read a lot of bank statements.