No user-submitted related servers found.
Scored across 27 tools
Descriptions do heavy lifting with explicit 'prefer X over Y' guidance, but four different read paths for a chat (chats_get, chat_outline, chat_read, chat_cache_read) plus chat_cache_sync/append/update_block create genuine boundaries an agent must reason through before every read. The explicit legacy duplicate containers_list == projects_list is a hard naming collision, and chats_create vs chat_ask also overlap.
Nearly all tools follow namespace_action (browsers_*, projects_*, jobs_*, chats_*), which is predictable and readable. Minor deviations: the chat singleton tools (chat_outline, chat_read, chat_ask, chat_cache_*) use a different prefix than the plural chats_* group, and containers_list sits outside the projects_* namespace it belongs to.
Twenty-seven tools across five real sub-domains (browsers, platforms, chats, cache, projects, jobs) is defensible given the scope, but it is above the comfortable range and forces the agent to triage before acting. At least one slot (containers_list) is pure dead weight, and the three browser tools plus six cache tools could be trimmed.
Coverage is unusually thorough: browser lifecycle, provider discovery, chat read/export/create/ask/delete, offline cache CRUD, project CRUD plus chat rehoming, and full async job management with cancel. Gaps are minor: no chats rename/update or metadata patch, no projects rename, and no job retry.