Free document toolsPOST /v1/warrants

Verify claim documents before you pay

Supported Formats: Repair Invoices, Parts Quotations, Trade Assessments
Core Verification Mechanisms: Line-Item Recomputation · ABN/Tax ID Check · Metadata Forensics
Need copy-paste code with sandbox testing?Open Developer Playbook (cURL / Python / MCP) →

The problem

Property and auto insurance claims are vulnerable to ghost repairs, altered quote totals, and non-existent subcontractor ABNs.

Claims handlers under pressure to meet turnaround targets rarely re-verify tax math or run ABN validity checks by hand.

How Stipple helps

Automated claim intake gate: verifies invoice arithmetic, checks repair quotes against supplier details, and inspects PDF rendering layers for spliced amount values.

Returns a clear triage flag: straight-through payment for clean claims, or an annotated finding list for the special investigations unit (SIU).

bash
curl -X POST https://www.stipple.sh/v1/warrants -F "file=@repair-invoice.pdf"
# → { "risk_band": "high", "recommended_action": "review_before_action",
#     "signals": [ { "id": "domain.invoice.reconciliation", ... } ] }

Protect loss ratios by catching fabricated invoices before cash disbursements are executed.

The evidencePayslip forgery benchmark: 210 documents, six forgery families

How the corpus was built, what was caught, and what the method does not prove.

Questions

What forgery types does it catch on invoices?
The benchmark families: arithmetic that does not recompute, statutory-rate deviations, spliced fonts, and identifier checksums that fail — plus metadata and tamper signals. A consistent forgery that adjusts every figure in step still reconciles, and the result says so rather than implying a clean bill.

Sources

  1. 01Payslip forgery benchmark: 210 documents, six forgery families
  2. 02Developer playbook: invoice line items and GST