Scored across 41 tools
Most tools target a clearly distinct resource+action (read_base vs query_base, get_backlinks vs get_outlinks, list_tags vs search_by_tag), and descriptions explicitly cross-reference when to use which. The one genuinely crowded area is content mutation — append_to_note, prepend_to_note, update_section, insert_at_section, edit_block, and replace_in_note all write into a note — plus four separate search modalities (search_notes, search_semantic, search_by_frontmatter, search_by_tag), which are well-described but still require careful reading to pick correctly.
Every name is snake_case and verb-led with a clear object (get_note, list_canvases, find_broken_links, add_canvas_edge, search_by_tag). Family prefixes (list_*, get_*, find_*, search_*, create_*, add_*, *_canvas_*) form a predictable and readable taxonomy with no camelCase or casing drift.
41 tools is well past the point where an agent can hold the surface in view comfortably, and several clusters (six note-mutation tools, four search tools, three attachment tools) could be consolidated behind parameters. The domains are genuinely broad (notes, sections, blocks, tags, links, graph, canvas, Bases, attachments, embeddings), so it is not arbitrary bloat — but the selection cost on every turn is real.
Note lifecycle is essentially full (create/read/append/prepend/move/delete, frontmatter, sections, blocks, regex replace) and tags, links, graph traversal, Bases, attachments, and semantic search all have coherent mini-surfaces. Gaps are around the peripheral formats: canvas support is add-only (no node/edge update or deletion, no canvas creation), Bases are read/query-only, and there is no attachment deletion tool — the description even defers that to the shell.