Is Grafana AI only available in Grafana Cloud?

No. The Grafana MCP server and AI SDK are open source and run wherever you run Grafana. Agent Observability instruments agents with an open source SDK. Grafana Assistant, Knowledge Graph, gcx, and Cloud MCP deliver their full experience in Grafana Cloud.

Explain AI in observability vs Observability in AI.

AI in observability applies AI to operate your systems better and refers to the use of AI as part of a larger observability strategy. This could include agents baked into a platform (e.g., Grafana Assistant in Grafana Cloud) or other integrations that help automate and accelerate the ways teams observe their systems. AI observability is the use of observability to track the state of an AI system, such as an LLM-based application. It’s a subset of observability focused on a specific use case, similar to database observability for databases or application observability for applications. Grafana Cloud does both: AI that helps you operate, and observability for your AI.

What's the difference between the MCP server, gcx, and Grafana Assistant?

The MCP server is the open bridge that connects any agent to Grafana. gcx is the Grafana Cloud CLI that wraps that surface with deterministic commands plus Assistant reasoning in your terminal. Grafana Assistant is the purpose-built observability AI you talk to. Use MCP or gcx to give your own agent context; use Assistant when you want observability reasoning ready out of the box.

Is it safe to run AI against my production data?

Your data stays where it lives. Grafana AI reasons across it rather than requiring you to move it into a proprietary store. In Grafana Cloud, AI runs inside Grafana Cloud's compliance boundaries with role-based access control, and Guards let you define what agents may do without asking. Grafana Cloud is FedRAMP, PCI DSS, and GDPR compliant and SOC 2 Type II verified.

AI for observability, observability for AI

Understand, investigate, and act with AI

Grafana Labs brings AI across your whole observability workflow: ask questions in plain language, follow the evidence to root cause, observe the AI agents you ship, and give your own coding agents real production context. Connect through the open source Grafana MCP server, build on an open source SDK, or run the full experience in Grafana Cloud. Your data stays where it lives.

Create free account Docs

Understand systems faster

Ask questions about production telemetry, build dashboards, and explain logs in plain language with Grafana Assistant

Find root cause automatically

Let AI follow the evidence across metrics, logs, traces, and deployments to surface what broke and why with Investigations

Ship AI agents you can trust

Replay, evaluate, and continuously improve the AI agents you run in production with Agent Observability

Trusted by everyone from startups to the Fortune 500

Why use AI in Grafana Cloud?

The same AI works across the whole observability lifecycle: grounded in your real telemetry, governed by the controls you already trust, and available everywhere your team works.

Ask anything, grounded in your telemetry

Onboard in minutes with chat-driven workflows
Automate operational workflows and recurring analyses
Connect your AI tools with MCP, Workspace, and Automations
Get step-by-step guidance with AI-assisted support
Use natural language to write queries and build dashboards

Learn about Assistant

Follow the evidence to root cause

Automatically correlate metrics, logs, traces, and deployments when an alert fires
Test each hypothesis against your production telemetry — prove it or eliminate it, no guesswork
Share a transparent report so your team resolves the incident together, without recreating the work

Learn about Investigations

Observe the AI agents you ship

Understand how production AI agents behave with visibility into conversations, tool calls, latency, cost, and quality
Replay every agent decision to debug workflows and understand how responses were generated
Run live evaluations to detect hallucinations, unsafe output, and quality drift before users notice
Optimize latency, token usage, and costs across agents, runs, and models
Protect users from unsafe output with Guards that enforce rules and block violations before they reach production

Learn about Agent Observability

Give every agent a live map of your services

Agents that always know your environment: the Knowledge Graph delivers live runtime context across services, dependencies, and infrastructure as your stack changes, so nothing is ever stale
Every agent shares the same map: yours or Grafana's, all reasoning over the same live context without rebuilding it from scratch each session.

Learn about the Knowledge Graph

The Grafana Cloud AI portfolio

Multiple products, one platform. Start with any of them on the free tier.

Grafana Assistant

Understand systems, investigate incidents, and take action in natural language. Generally available in Grafana Cloud; in public preview for self-managed Enterprise and OSS.

Explore Grafana Assistant

Investigations

AI-powered root cause analysis that auto-investigates telemetry and follows the evidence to resolution.

Explore Investigations

Agent Observability

Evaluate AI agents in production. Replay conversations, trace tool calls, track costs, and catch unsafe output, hallucinations, and policy violations.

Explore Agent Observability

gcx + Cloud MCP

The official Grafana Cloud CLI: deterministic commands plus Assistant reasoning in your terminal and your coding agents. Generally available.

Explore GCX

AI in Grafana OSS

Grafana AI is open at the foundation. The building blocks (the MCP server, the AI SDK, and the LLM plugin) are all open source, so any team can connect an agent to Grafana, build their own, or add AI to a self-managed stack. Start where your data already runs, and scale into Grafana Cloud when you're ready.

Connect an agent, or build your own

AI SDK: Build and instrument your own agents and AI applications on an OpenTelemetry-native Go toolkit.
MCP server: Connect Claude Code, Cursor, or any MCP client to Grafana for grounded, real-telemetry context (self-managed or Cloud)
OSS LLM plugin: Bring your own model provider and keys, and add explain-in-place across your dashboards
o11y-bench: Grafana's open-source benchmark, lets anyone test any AI agent on real observability workflows because you should be able to see how agents actually perform before putting them in production

The Grafana OSS AI portfolio

Not all of Grafana Labs AI products are Cloud-only. Bring AI into the workflows you already run, open by default.

Grafana MCP server

The open source bridge that connects any agent (Claude Code, Cursor, Copilot) to Grafana, whether self-managed or Cloud.

Explore the Grafana MCP server

AI SDK

A backend toolkit for building AI-powered applications and agents. It’s a Go SDK for building your own agents, with support for Grafana Cloud, Anthropic, OpenAI, Amazon Bedrock, and any OpenAI-compatible endpoint.

Explore the AI SDK on GitHub

Grafana Assistant

Understand systems, investigate incidents, and take action in natural language. In public preview for self-managed Enterprise and OSS; also in Grafana Cloud.

Explore Grafana Assistant