No user-submitted related servers found.
Scored across 28 tools
Each tool maps to a distinct resource/action; the CI hierarchy (run, workflow, job, attempt) is cleanly separated across get/list/diagnose tools, and risky pairs like builds vs CI runs or diagnose vs logs are explicitly cross-referenced. No two tools appear to serve the same purpose.
The set follows a consistent depot_<verb>_<object> snake_case convention, with list/get/diagnose verbs and clear resource nouns. Minor deviations like depot_whoami and depot_wait_for_ci_run break the pure verb_noun pattern but remain predictable.
With 28 tools this crosses the 'too many' threshold and creates a heavy selection surface for an agent, even though the Depot API scope is broad. Splitting this into CI, build/registry, and admin/usage servers would make the surface easier to navigate.
The read-only observability domain is covered very well: diagnosis, comparison, waiting, logs, metrics, artifacts, job summaries, build diagnostics, cache, usage, tokens, trust policies, and secret/variable scoping are all present. The main gap is the lack of mutating tools (rerun, retry, cancel, delete), which descriptions reference but do not expose; this is a deliberate read-only design rather than a fatal omission.