Skip to main content
Watch: I Built an AI App Without Leaving Databricks

Observability - official Grafana server

Grafana MCP server

Open source, self-run MCP server for any Grafana instance (Grafana Cloud or self-managed) that exposes tools for dashboards, datasources, Prometheus, Loki and other queries, alerting, incidents, OnCall, annotations and admin. Authenticates with a Grafana service account token or username and password. It runs locally from mcp-grafana (pypi) with API key auth.

Last verified 15 tools7 sources

Install

VS Code

JSON
"mcp": {
  "servers": {
    "grafana": {
      "type": "sse",
      "url": "http://localhost:8000/sse"
    }
  }
}

Generic

JSON
{
  "mcpServers": {
    "grafana": {
      "command": "uvx",
      "args": ["mcp-grafana"],
      "env": {
        "GRAFANA_URL": "http://localhost:3000",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<your service account token>"
      }
    }
  }
}

Tools

search_dashboards
Search for dashboards by query, folder UID, tag, or starred.
get_dashboard_by_uid
Get a dashboard by uid, optionally a saved version.
update_dashboard
Update or create a new dashboard.
get_dashboard_summary
Get a compact summary of a dashboard without full JSON.
run_panel_query
Execute one or more dashboard panel queries.
list_datasources
List datasources.
check_datasources_health
Check the health of one or more datasources.
create_datasource
Create a datasource.
query_prometheus
Execute a query against a Prometheus datasource.
list_prometheus_metric_names
List available metric names.
query_prometheus_histogram
Calculate histogram percentile values.
list_incidents
List incidents in Grafana Incident, optionally with their custom field values.
create_incident
Create an incident in Grafana Incident, optionally setting custom fields.
list_teams
List all teams.
user_info
Current identity, capabilities, and accessible organizations.

As documented on October 17, 2026. A running server reports its current list through tools/list.

Notes and caveats

  • Requires Grafana 9.0 or later for full functionality.
  • Latest release 2.0.2 (CHANGELOG dated 2026-10-07; also on PyPI and Docker Hub).
  • The Docker image defaults to SSE mode; pass -t stdio for stdio clients.
  • Networked modes (SSE, streamable-http) should set MCP_GRAFANA_SERVER_TOKEN for caller auth.
  • GRAFANA_API_KEY is deprecated in favour of GRAFANA_SERVICE_ACCOUNT_TOKEN.
  • Also deployable via the grafana/grafana-mcp Helm chart.
  • The quick-start JSON is documented for Claude Desktop and Cursor.
  • The VS Code snippet assumes the server is running locally in SSE mode.
  • No preview or GA label in the docs, so status is left unknown.

More MCP servers for observability

Related reading

FAQ

What does the Grafana MCP server do?

Open source, self-run MCP server for any Grafana instance (Grafana Cloud or self-managed) that exposes tools for dashboards, datasources, Prometheus, Loki and other queries, alerting, incidents, OnCall, annotations and admin. Authenticates with a Grafana service account token or username and password. The vendor documents 15 tools, including search_dashboards, get_dashboard_by_uid, update_dashboard and get_dashboard_summary.

How do I add the Grafana MCP server to VS Code?

Grafana documents a VS Code config; copy it from the VS Code section of this page. Snippets for Generic are there too.

Does the Grafana MCP server need authentication?

Yes, API key. Transport: stdio, SSE, Streamable HTTP. Status: Unlabelled, last checked October 17, 2026.

Sources

Last verified against 7 Grafana pages, then re-checked independently.

Newsletter

New MCP servers, the week they ship

One email per week. New official MCP servers, the coding tools that support them and how developers wire them up.