This is documentation for the next version of Grafana documentation. For the latest stable release, go to the latest version.

AI tools

Grafana Cloud Enterprise Open source

AI tools in Grafana OSS

If you’re a self-managed Grafana user, you have the following AI tools available to interact with Grafana:

Grafana Assistant is the Grafana AI agent, built into Grafana. Assistant brings a purpose-built LLM into Grafana so operators, developers, and SREs can troubleshoot issues, manage dashboards, and answer product questions in plain language.

In self-managed Grafana the Assistant app runs in your own deployment but connects to a Grafana Cloud Assistant backend, with a reduced feature set.
Assistant is billed per token consumed, including requests made through gcx and MCP servers (although Assistant in MCP is disabled by default).

MCP servers let an external AI client (for example Claude, Cursor, or VS Code with Copilot) call Grafana as a set of tools that cover common tasks: dashboards and folders, data sources (Prometheus, Loki, and others), alerting, incidents, OnCall, and more.

A model picks which tool to use based on the tool description, and calls the tool with structured arguments. Grafana runs it under normal identity and RBAC, and the structured result goes back to the model for its next step.
Grafana offers two servers: Cloud MCP (hosted by Grafana) and OSS MCP (you run it).

gcx is an AI-native command-line tool that gives you or your AI agent structured access to Grafana (dashboards, folders, alert rules, data sources) and Grafana Cloud products such as Synthetic Monitoring, k6, Fleet Management, Incidents, and Adaptive Telemetry. Use it to script deterministic JSON or YAML output, defined errors, and predictable exit codes.

Explore

Grafana Assistant

Install Grafana Assistant in self-managed Grafana and connect it to a Grafana Cloud stack.

MCP server

Connect AI assistants and LLM clients to Grafana using the Model Context Protocol.

gcx

Use gcx, a command-line tool you can run yourself or hand to an AI agent, to authenticate, manage multiple environments, and perform administrative tasks from the terminal.

Side-by-side comparison

This table summarizes the main points for each tool:

	Grafana Assistant	Cloud MCP	OSS MCP	gcx
What it is	Built-in Grafana AI agent	Grafana-hosted MCP endpoint for external AI clients	Self-run MCP server (`mcp-grafana`)	CLI for Grafana and Grafana Cloud
Who drives it	A person, inside Grafana	A person or an agent, through an MCP client	A person or an agent, through an MCP client	A person, a script, or a coding agent issuing explicit commands
Setup	In Cloud, there’s nothing to install, just enable it in your stack. Install the plugin for self-managed	Add one endpoint to a client; complete browser consent	Install and operate a server	Install the binary and configure contexts
Best for	Specific questions, dashboards, investigations in the UI	Interactive diagnosis from your AI client with per-user permissions	Self-hosted, private networks, service identity, custom policy	Repeatable work, CI, GitOps, multi-stack changes
Cost note	Per token consumed	Counts toward Assistant usage and active users	Server hosting is yours. Grafana usage rules apply	Free to use. Grafana usage rules apply.

When to use each tool

See the following scenarios to better understand when to use each tool:

I…	Use	Why?
Want an answer, a query, or a dashboard fast from the Grafana UI	Assistant	No setup, and Grafana manages context.
Want to allow my LLM to interact with my OSS Grafana instance for common tasks	OSS MCP	`mcp-grafana` has a suite of tools for common tasks.
Don’t have an AI subscription, but I want to interact with my self-managed Grafana instance using natural language	Grafana Assistant on-prem	It lets you bring the power of Grafana Assistant to your self-hosted Grafana instance.
Have CI pipelines, GitOps, dashboards as code, or work across several stacks	`gcx`	Explicit commands, exit codes, and named contexts make changes repeatable and reviewable.
Want conversational diagnosis followed by a reviewed change	MCP for diagnosis, `gcx` for the change	Refer to Use them together.
Use a coding agent needs flexibility to build its own workflows	`gcx`	`gcx` is as a broader, less opinionated toolset.

Use the AI tools together

You may use a composite pattern for coding agents: for example, use MCP for conversational discovery and diagnosis, then move durable changes into explicit gcx commands and version-controlled files.