GroundTruth: 2026 US Tariff & Landed Cost

compute_tariff performs a calculation while list_tariff_categories returns reference data—two clearly distinct purposes with no overlap. An agent can trivially pick the right one based on whether it needs an estimate or raw category information.

Both tools follow a clean verb_noun snake_case pattern (compute_tariff, list_tariff_categories). The verbs match their actions precisely and the noun structure is parallel.

Two tools is thin for a domain that involves categories, country layers, fee types and computed estimates. It is defensible as a focused utility, but one or two more tools (e.g. category lookup or rate detail) would feel less sparse.

The core workflow—list the tariff landscape, then compute landed cost from a description—is covered, and compute_tariff returns sources and as-of dates. Gaps are minor, such as drilling into a single category or comparing multiple origins in one call.