Monitor MCP Server

Valkey and Redis investigations in BetterDB can begin with selecting an instance and checking health, memory, client connections, and replication. The monitor tools let an assistant move from recent slow commands to persisted patterns, latency histories, and access audits when investigating an incident. Cluster views combine node performance, aggregated slowlogs, and slot statistics, while hot-key snapshots show uneven access. Capacity forecasts estimate when a metric may reach its resource ceiling. Anomaly detection requires BetterDB Pro, and COMMANDLOG analysis and slot statistics require Valkey 8+.

For AI workloads, semantic caches expose similarity scores and threshold recommendations, while agent caches report per-tool effectiveness and TTL recommendations. Cache adjustments and invalidations are handled as review proposals, with approval and rejection tools; threshold and TTL changes require human approval. Agent-memory tools discover stores, browse stored items, report statistics, and recall memories using an embedding supplied by the caller. Deletions likewise require human approval. OpenTelemetry trace tools expose span timing and attributes, then correlate cache and memory spans with live Valkey state on a selected instance. Stored component metrics provide a separate history of cache and memory behavior.