Each tool targets a distinct resource-action: searching entities (search_nodes), listing episodes (get_episodes), searching facts (search_memory_facts), fetching a single edge (get_entity_edge), tracing provenance (get_episode_entities), and health (get_status). Mild overlap exists between search_memory_facts and get_entity_edge (both about fact edges) and between search_nodes and get_episode_entities (both surface entities), but descriptions differentiate them adequately.

All names follow a clean snake_case verb_noun pattern (search_*, get_*). Minor inconsistency: only search_memory_facts carries the 'memory' qualifier while its sibling search_nodes targets entities, and the get_* prefix is overloaded across heterogeneous resources.

Six tools is a reasonable, well-scoped set for a graph-memory server, with each tool covering a distinct retrieval need. It leans slightly thin—no ingestion or mutation tools—but the read surface is coherent.

The surface is entirely read-only: no operations to add episodes/entities/facts, update them, or delete them, which would normally be part of a knowledge-graph lifecycle. As a local read consumer of an externally built graph this may be intentional, but the missing write/traversal operations are a notable gap.