Related Servers

The descriptions are unusually careful about cross-referencing close pairs (e.g. get_course_grades vs obs_get_course_grades, get_course_attendance vs obs_get_attendance), which helps a lot. However, across 108 tools there are many genuinely overlapping surfaces: get_dashboard vs get_dashboard_announcements/assignments, get_course_overview vs every individual get_course_* reader, get_upcoming_deadlines vs the dashboard assignment readers, and a dense cluster of archive_* and library_* readers that an agent can easily misselect between. This is squarely 'some overlap exists but descriptions help'.

Almost everything is snake_case with a namespace prefix (obs_, archive_, library_, mail_, get_course_, etc.), which is readable and generally predictable. But the convention is not uniform: some names are verb_noun (get_dashboard, read_page, search_campus_locations), others are bare noun_status forms (auth_status, mail_status, obs_auth_status, archive_status), and verbs vary widely (get/read/list/search/crawl/snapshot/diff/watch/explain/build/plan/estimate). Mixed but still decipherable.

108 tools is an extreme mismatch for any single server surface, far beyond the 50+ threshold that indicates poor scoping. Even accounting for the multi-domain scope (Ninova, OBS, library, archive, mail, campus), the set is bloated with overlapping readers and four known-dead legacy library tools, forcing the agent to wade through a huge undifferentiated list to find the right call.

Coverage is remarkably broad and deep: Ninova LMS reads/writes, OBS records, registration draft/validation, grades, transcripts, attendance, prerequisites, library catalog, mail, calendar, shuttle, cafeteria, help tickets, directory, and a historical course archive with seasonality and fill-rate analysis. The main gaps are the four legacy library tools that no longer function (account, loans, renew, hold) and the absence of any real OBS add/drop execution, which agents can work around via the official pages.