Built to read anything, and to keep up.
Customs compliance is a moving target with messy inputs. The platform is designed around both facts.
Structured or not, we read it.
Entry summaries, invoices, packing lists, bills of lading, consignment sheets, broker statements: typed, scanned, photographed or exported. TariffSense understands the layout, extracts the lines, and matches them across documents. New layouts are supported quickly because the reading is done by AI, with deterministic checks where precision matters.
Consignment sheet
A
B
A
C
B
Broker statement
Photo_0412.jpg
Lines across documents3 lines
| Description | HTS code | Qty | Value |
|---|---|---|---|
| Cotton knit shirtsSheetStmtPhoto | 6109.10.0012 | 1,200 pcs | $12,000 |
| Steel shelving bracketsSheetStmt | 7326.90.8688 | 400 pcs | $8,000 |
| Aluminum plateSheetStmt | 7606.11.3030 | 2,000 kg | $10,000 |
| Total | $30,000 | ||
Tariff changes, captured within hours.
When a tariff notice is published, our AI-assisted pipeline captures it within hours and ships the tested rules within days. Every rate is stored with the dates it applied, so historical entries are always judged by the rules of their day.
Your exports in. Your reports out.
Bring data the way it already leaves your systems: broker entry exports, ERP billing and invoice exports, carrier manifests, spreadsheets. Get back audit reports, corrected values, filled forms, and programmatic access through our API and MCP server for your own tools and assistants.
Separate by client, business unit or brand.
Brokers and multi-entity importers keep each client's products, entries and findings in their own workspace, with roll-up views when you need the whole book.
Every entry learns from the last one.
Findings look beyond the single entry: the same product classified differently last quarter, a value that drifted from the usual range, an origin that changed without a reason. Patterns a sample audit would never see.
Your documents stay yours.
TariffSense runs on Google Cloud. Your documents are encrypted in transit and at rest and scoped to your company's workspace. We don't train models on your data, and neither do the model providers we use. Your files are used for one thing: checking your entries.
See the platform on your own documents.
Bring your messiest packet. We'll read it live.