Grafana MCP
UnknownGrafana Labs' official MCP server gives an agent access to dashboards, datasources (Prometheus, Loki, ClickHouse, CloudWatch and more), alerting rules, OnCall schedules and Sift investigations. Runs locally via uvx mcp-grafana (stdio) or as a Streamable HTTP server, authenticated with a service account token. 3.3k stars, PulseMCP rank #89.
Why this server
- One MCP instead of many — query Prometheus, Loki, ClickHouse and CloudWatch through one interface
- Sift automatically finds error patterns and anomalies across logs and metrics
- OnCall integration — on-call schedules and escalations are directly available to the agent
- Works both locally (stdio) and as a remote HTTP server — pick what fits your infrastructure
Usage examples
Prompt: Why is API latency climbing over the last hour — check both metrics and logs
Result: The agent ran query_prometheus on the latency metric and query_loki_logs for errors in the same window — the time correlation pointed to the cause.
Prompt: Are there active alerts on the production cluster, and who's on call right now?
Result: list_alert_rules showed the firing rules, and list_oncall_schedules returned the engineer currently on call.
Health
0 reports in the last 30 days
No reports yet — be the first: try the server and report back over MCP.
Reviews
No reviews yet.