madda — Moroccan law, article by article

Each tool targets a distinct resource+action: article body (get_article), text metadata/TOC (get_text), article history (get_versions), taxonomy (list_categories), and retrieval (search_law). The only mild adjacency — get_article vs get_versions, since the latter also returns current wording — is explicitly disambiguated in the descriptions. No two tools would plausibly be misselected for the same intent.

All five names follow a strict verb_noun snake_case pattern (get_*, list_*, search_*). Verb choice is semantically consistent: get_ for retrieval of a known item, list_ for enumeration, search_ for querying. No mixed conventions or stray casing.

Five tools is a lean but coherent surface for a read-only legal corpus: search, browse categories, text metadata, article, and version history. It sits at the low end of the ideal 3-15 range and one more browsing tool would make the set feel complete, but nothing is redundant.

The core read path is fully covered: find a law (search or categories), get its TOC, then fetch articles and their historical wordings with citations. The main gap is that list_categories returns text counts but there is no tool to enumerate the texts within a category — agents must fall back on search to drill down. Read-only scope means no create/update/delete is expected, so this is a minor, workable gap.