CourtWatch Agent API

Each tool targets a distinct stage of retrieval: about gives domain/plan/document-id metadata, search finds passages from a natural-language question, and fetch reads a single document by id. The about/search boundary is slightly soft since about also surfaces document ids and content, but the descriptions make the intended flow clear.

All names are short, lowercase, unprefixed single words, which is a consistent and readable convention. The only minor deviation is that fetch and search are verbs while about is not, but the style remains uniform.

Three tools is on the lean side, but for a read-only public-content retrieval API that is a well-scoped surface with no redundant entries. Nothing feels missing at the count level, though a fourth tool (e.g. list/browse) could be justified.

The discover-search-fetch lifecycle is fully covered for a read-only documentation/retrieval domain, and write operations are not applicable. Minor gap: no way to enumerate documents or browse by site/date beyond issuing a search query.