Netdata MCP

Netdata's MCP server gives an AI client access to per second infrastructure telemetry, which is what separates it from monitoring integrations that only expose coarse dashboards. Discovery tools come first: node discovery returns hardware specifications, operating system details and versions for connected nodes; metrics discovery runs full text search across contexts, instances, dimensions and labels so an agent can find the right chart without knowing Netdata's naming; alert discovery reports which alerts are raised right now; and function discovery lists the live functions a node supports, including processes, network connections, streaming state, systemd journal and Windows events. Query tools then do the analysis. Metrics queries support complex aggregations with machine learning anomaly detection applied, and metrics scoring performs root cause analysis by combining anomaly rates with metric correlations, so an agent can rank which of hundreds of charts actually changed when an incident started. Alert history returns the full transition log for an alert, function execution runs a function on a connected node, and log exploration reads logs from those nodes. There are two ways in. A single cloud hosted endpoint serves every node in a Netdata Cloud space and is authenticated with an API token carrying the MCP scope. Alternatively a client connects straight to any Netdata Agent or Parent on the network using its locally generated API key, keeping data on the customer's infrastructure. It fits DevOps engineers and system administrators doing AI assisted troubleshooting.