Related Servers

The set contains many functionally overlapping families: ecosystem_scan / ecosystem_scan_periodic / ecosystem_refresh / ecosystem_index_update all trigger scans with subtly different semantics, and ecosystem_summary_top_n / _health / _by_tag / _weekly are near-identical summary readers. Task, memo, memory and briefing tools likewise blur boundaries (task_memo_read vs task_execution_trace vs unified_search; memory_list vs memory_search vs memory_reconcile_candidates). Descriptions are unusually long and do disambiguate, but relying on multi-paragraph prose to separate 100+ tools means misselection is likely.

Almost everything is snake_case with a domain prefix plus action (project_create, team_list, meeting_conclude, channel_send), which is a clear, predictable pattern. Deviations are minor and mostly historical: a few verb-first or prefix-less names (unified_search, find_skill, verify_completion, decision_log, event_list, usage_attribution) break the domain_verb shape. No camelCase/snake_case mixing.

116 tools is an extreme over-extension of the surface; the rubric treats 50+ as a 1. Whole families exist only because functionality was never consolidated (scan vs scan_periodic vs refresh vs index_update; four summary endpoints; two task-wall entry points folded from older tools). The count makes discovery and correct selection impractical for an agent.

The surface is broad, covering projects, tasks, teams, agents, meetings/debates, memory layers, ecosystem pipeline stages, reports, workflows, channels and fleet dispatch. But core lifecycle gaps remain: there is no team_create or agent_create (both appear auto-generated), no task_delete, and memory deletion is only approximated by invalidate. Agents can work around these, but an 'OS' with 116 tools should not have missing create/delete on its central entities.