No user-submitted related servers found.
Scored across 12 tools
Most tools have distinct read/action targets: search_knowledge (browse) vs get_knowledge (detail) vs family_tree (graph walk) vs knowledge_signals (usage) vs teacher_profile (provider) are separable by description, and the job_* trio splits cleanly into status/list/cancel. The main blur is quote vs get_knowledge (both take a knowledge id and both report stack/ownership) and my_library vs node_status, which overlap on node-local holdings and capabilities.
Conventions are mixed: verb_noun (search_knowledge, get_knowledge, get_training_set), noun_status compounds (job_status, node_status), and bare nouns (quote, family_tree, my_library, knowledge_signals, teacher_profile). Everything is snake_case and readable, but there is no predictable pattern an agent can rely on to guess a name.
12 tools sits in the healthy range and each of the twelve occupations (library, job status/cancel/list, quote, search, get, family tree, training set, node status, signals, teacher profile) is plausibly needed. It is slightly heavy on read-only inspection tools relative to the number of mutating operations actually exposed.
The descriptions repeatedly reference tools that do not exist in this surface: buy (explicitly required to follow quote), live_test, teach, teach_preflight, apply_knowledge, remove_knowledge, create_training_set, and reconcile_purchase. An agent can quote and inspect, but cannot actually complete the purchase/teach/live-test lifecycle the server is built around, leaving dead ends at the key decision points.