Auto-discovers MCP servers from your Claude Desktop, Cursor, or VS Code configs and probes them by calling list_tools, the same handshake your agent uses. Unlike HTTP uptime monitors that just ping a port, this verifies the server actually speaks MCP and tracks tool schema drift across versions. Stores latency history in local SQLite with p50/p95 percentiles, generates self-contained HTML dashboards with uptime sparklines, and flags degraded servers based on response time thresholds. Runs as a daemon for continuous polling or on-demand via eight tools covering health checks, history queries, manual server registration, and update detection. Reach for this when you need protocol-aware monitoring without shipping data to external services.
claude mcp add --transport stdio dbsectrainer-mcp-server-health-monitor -- npx -y mcp-server-health-monitor