Tango

Descriptions are unusually explicit about boundaries — add_comment vs add_progress_note, log_project_decision vs memory_save, find_people vs resolve_mention are all carefully delineated. However, with 99 tools there are still several near-neighbors (accept_task vs complete_task; update_project_decision vs log_project_decision; the many list_project_* readers) that could trip an agent without reading the fine print.

The set overwhelmingly follows a predictable verb_noun snake_case pattern (list_tasks, create_task, get_client_context, update_worker, rotate_webhook_secret). A few outliers (whoami, glossary, check_in) break the pattern but are few and readable in context.

99 tools for one server is a heavy surface that forces agents to navigate a large mental model and increases selection cost. Even a rich task/workflow platform rarely needs this many exposed operations, and the count sits well past the '25+ feels too many' threshold.

Coverage is exhaustive across the domain: full task lifecycle (create/update/claim/lease/handoff/pause/complete/accept/delete), worker identity and key management, client/project/context/decision/event/issue CRUD, memory, integrations, webhooks, and even cryptographic transparency verification. Few obvious gaps.