No user-submitted related servers found.
Scored across 33 tools
Most tools target clearly distinct resources and actions (capture vs save_source vs write_page vs read_page), and descriptions supply explicit disambiguation cues like 'use this instead of capture when the content has a link'. Residual confusion lives in the review-queue family — list_pending, list_pending_revisions, list_pending_imports, list_refinements, plus confirm_memory/accept_revision/accept_refinement — and in the near-twin page_basis/page_code_basis pair, where an agent must read carefully before choosing.
The dominant pattern is snake_case verb_noun (list_pages, get_page_sources, delete_page, dismiss_revision, create_relation), which is applied consistently across CRUD and review operations. Deviations are minor: a handful of bare verbs (capture, recall, distill, lint, forget, brief) and two noun-first names (page_basis, page_code_basis).
33 tools is on the heavy side, and a tail of rarely invoked, local-stdio-only operations (lint_repair_prepare/apply/verify, page_code_basis, list_rejections, list_pending_imports) inflates the surface. The domain genuinely spans several subdomains — memory lifecycle, distillation, wiki pages, lint/repair, refinements, briefs — so each tool has a defensible role, but the set is borderline oversized for a single MCP endpoint.
Coverage is unusually thorough: memories have capture/recall/confirm/forget plus revisions, pages have full create/read/list/write/rename/delete with revisions and sources, and review queues, lint, briefs, entities, and relations all have entry points. Gaps are minor but real — no direct get_memory-by-id (forget depends on finding the memory via recall first), no entity/relation listing or deletion, and no page search beyond list_pages paging.