site

All tools target the same single dataset, and the three row-retrieval tools (dataset_row, dataset_search, dataset_compare) have overlapping purposes that could cause misselection. Each has a distinct matching semantic (exact, contains, multi-value), but the boundaries are subtle and rely heavily on reading descriptions carefully. Dataset_columns and dataset_provenance are clearly distinct from the rest.

Every tool follows a consistent dataset_<suffix> pattern, which makes the set predictable and easy to scan. The suffix grammar is mixed—nouns (columns, provenance, row, stats), verbs (compare, search), and an adjective (top)—so it is not a uniform verb_noun convention, but the prefix consistency keeps it coherent.

Seven tools is well-scoped for a single-dataset query assistant and falls comfortably in the ideal range. Each tool earns its place by covering a distinct query type: schema discovery, provenance, exact lookup, substring search, multi-value comparison, numeric statistics, and top-N ranking.

The read-only surface covers the core analytical workflows well: schema discovery, attribution, exact and fuzzy lookup, comparisons, statistics, and ranking. Minor gaps exist—such as grouped aggregation or listing distinct column values—but agents can work around these with the provided search and row tools.