Can AI agents automate Grafana?

Grafana is automatable today through a public HTTP API and a first-party MCP server. Every Grafana instance exposes an HTTP API that the product frontend uses to save dashboards, create users, update data sources, and delete alerts. Grafana 13 deprecates legacy /api routes in favor of Kubernetes-style /apis endpoints; legacy routes remain available but are no longer updated, and Grafana says new APIs may not match every legacy path yet, so coverage is partial. Grafana Cloud authenticates the HTTP API with a service account token only. Grafana OSS and Enterprise also accept HTTP basic authentication. Official OpenAPI v2 and v3 specifications describe the same route set. The official open-source Grafana MCP server at github.com/grafana/mcp-grafana connects MCP clients to Grafana OSS or Cloud with a service account token and exposes tools for dashboards, datasources including Prometheus and Loki, alerting, Grafana Incident, OnCall, and deeplinks; Grafana 9.0 or later is required for full functionality, and several query-tool categories are disabled by default. Grafana Cloud also hosts a remote MCP server at https://mcp.grafana.com/mcp using OAuth 2.1 and Streamable HTTP. That hosted server requires Grafana Assistant terms, the Assistant Cloud MCP User permission, and counts toward Assistant usage; write tools are optional at consent time and OAuth sessions last 30 days. The official Foundation SDK defines dashboards as typed code in Go, TypeScript, Python, PHP, and Java. The gcx CLI manages Grafana 12 or later resources as code and is documented for AI agents, but Grafana marks it as under continuous development and some Cloud commands are usage-billed; the bundled grafana cli remains a plugin and admin tool. Dashboards export and import as JSON. Alerting supports webhook contact points. n8n ships a Grafana node for dashboards, teams, and users. Zapier stated Grafana is not available. Microsoft's connector catalog had no Grafana connector on 2026-08-22. apps.make.com/grafana returned 404. Computer-use viability is unassessed. Loki, Tempo, and Mimir are not scored as standalone products; they appear here only as Grafana datasources or proxied MCP tools. API and MCP both score 8; API is the best path because it does not require a local MCP process or Grafana Assistant entitlements.

Best path today: api · Overall automatability: 8/10