MCP & Integrations
MCP connects Grafana to your tools both ways. Bring Grafana context to the tools where work happens, and connect the MCP servers your team already runs back into Grafana Assistant and your other agents.
The Grafana MCP server lets MCP-compatible agents, editors, and custom tools ask Grafana and act with context: a structured path into Grafana data and a structured path back with the work they produce.
- Bring metrics, logs, traces, dashboards, and service context into MCP-compatible clients.
- Let agents and tools pass structured findings, prompts, and actions back into your Grafana workflow.
- Compose an agent setup around your team's tools instead of forcing one fixed interface.
The Grafana MCP server is open source and developed in the open on GitHub. Grafana Cloud is an open agent platform you can drive from the coding agent you already use, not a replacement for it. Pick the path that fits your setup: gcx for coding agents, or the Grafana MCP server for any MCP-compatible client.
Connect MCP servers to Assistant and your agents
MCP is not one-directional. Grafana Assistant and your other agents can act as MCP clients too, so you can connect the MCP servers your team already runs and give an investigation the outside context it needs.
The killer use case: point Assistant at a GitHub MCP server and let it read your code during an investigation: correlate a spike or a firing alert with the change that shipped, the pull request that touched the service, or the function on the hot path, without leaving Grafana.
- Connect any MCP server (source control, issue trackers, internal tools) to Assistant, gcx, and your agents.
- Ground investigations in context that lives outside Grafana, like the code and pull requests behind a regression.
- Keep your Grafana identity and permissions in force wherever those agents reach.
Other ways to integrate
Working agent-to-agent, or wiring Grafana into your own services? Beyond MCP, Grafana AI also exposes A2A, webhooks, HTTP endpoints, and a frontend SDK.