Scored across 30 tools
Every tool targets a distinct resource+action combination. Similar verbs are separated by resource prefix (card_create vs memory_add vs doc_create vs changelog_add), and similar actions on the same resource are disambiguated by purpose: patch vs write vs note on cards, release vs preview on changelog, supersede vs patch on memory. An agent can reliably select among the 30 tools.
All tools follow a strict project_<resource>_<action> pattern. Resources are consistent (card, memory, changelog, doc) and verbs are consistent (create, list, patch, get) throughout. Even cross-cutting tools fit the pattern (project_search, project_next, project_doctor, project_workspace). No mixed conventions or stylistic deviations.
At 30 tools, the surface is heavy. While the resource families are each well-scoped, the total exceeds the 15-25 comfortable range and approaches the 25+ threshold. Several tools feel granular (project_changelog_preview vs project_changelog_list; project_memory_graduate vs project_memory_supersede; project_card_note vs project_card_write) and could plausibly be consolidated, suggesting this could be tightened to the low-to-mid 20s.
The server provides full lifecycle coverage across multiple domains: cards (create, list, get, claim, transition, patch, write, note, release, archive, reopen), memory (add, list, patch, graduate, supersede), changelog (add, list, patch, preview, release), docs (create, list, move, patch), plus cross-cutting utilities (search, next, doctor, workspace, context bundle). No obvious gaps in the canonical project-management workflow, and dead-end operations are paired with inverses (archive/reopen, claim/release, supersede).