Pane

Each tool targets a distinct resource or action, and the descriptions go to great lengths to delineate boundaries (e.g. add_row vs update_row, deploy_app vs document_upload, list_rows vs get_row vs count_rows). The main risk is the credential cluster (key vs credentials vs agent vs connections) and the publishing cluster (community vs publisher vs review), where an agent must read carefully to pick the right one, but no two tools truly do the same thing.

All names are snake_case, which is consistent, but the verb style is mixed: row/row-adjacent tools use verb_noun (add_row, get_row, delete_row, list_rows) while many others are bare nouns (agent, apps, attachments, connections, credentials, members, publisher, review, transfer). Convention is readable but not predictable—an agent cannot infer the pattern from a name alone.

At 27 tools the surface is on the heavy side of the stated 3-15 sweet spot and crosses the 25+ threshold. It is partly mitigated because multiple actions are grouped under single tools (apps, attachments, credentials, community), so this is defensible for a full app platform rather than gratuitous, but it is still a large set to navigate.

Coverage is unusually thorough: rows have full create/read/list/count/update/soft-delete/restore/list-deleted; apps have the whole lifecycle including audit, share links, domains, wake, soft-delete/restore/purge; and attachments, credentials, connections, members, grants, ingest, community publishing, reviews and transfers are all represented. No obvious dead ends or missing operations for the stated domain.