Part VI: Monitoring and observability

All observability in memo-mcp is local and pull-based. Metrics are served on loopback, logs go to stderr, and the optional call log is a table in your own file. Nothing is pushed anywhere, and there is no telemetry.

• Signals overview: what exists, where, and the per-process caveat.
• The metrics endpoint: enabling /metrics on the server and the UI.
• Metric catalogue: every series, its type, labels and meaning.
• Prometheus: scrape, rules, alerts: ready-to-adapt configuration.
• Logs: format, level, fields, and where they end up.
• Call and query logs: per-call history inside the knowledge base.