Bring Grafana Assistant to Your Self-Managed Grafana: A Step-by-Step Setup Guide
Introduction
Grafana Assistant, the purpose-built LLM agent, is no longer limited to Grafana Cloud. As announced at GrafanaCON 2026, you can now connect this powerful AI tool to your self-managed Grafana Enterprise or OSS instances. This means you can analyze telemetry data, build dashboards, investigate incidents, and more—without migrating to the cloud. Here's how to set it up step by step, with customization options to fit your specific needs.
What You Need
- A self-managed Grafana instance (OSS v10.0+ or Enterprise v10.0+)
- Internet access from that instance to Grafana Cloud
- A Grafana Cloud account (free tier available with generous limits)
- Admin privileges on your Grafana instance to install plugins
- Familiarity with natural language queries (optional but helpful)
Step-by-Step Setup
Step 1: Create a Grafana Cloud Account
Go to grafana.com and sign up for a free Grafana Cloud account. The forever-free plan includes Assistant with reasonable usage quotas, so you can start right away. You'll need this account to bridge your self-managed Grafana with Assistant's cloud-based AI engine.
Step 2: Connect Your Self-Managed Grafana
In your Grafana Cloud account, navigate to the Connections section and choose Self-Managed Grafana. You'll be given a one-click setup command or configuration token. Copy it. Then, in your self-managed Grafana instance, go to Administration > Plugins and data > Plugins and paste the token into the connection dialog. This establishes a secure tunnel—your raw data stays local; only processed summaries and results are transmitted.
Step 3: Enable the Grafana Assistant Plugin
After connection, install the Grafana Assistant plugin from the Grafana plugin catalog. Enable it, and configure it to appear in your navigation. You may need to restart Grafana to activate the plugin. Once done, you'll see a chat icon or tab in your interface.
Step 4: Customize and Start Using Assistant
Click the Assistant icon to open the chat panel. You can now ask questions in natural language, like "Show me all error logs from the last hour" or "Create a dashboard for CPU usage across hosts." The assistant uses your telemetry data to answer. For advanced customization, you can fine-tune which data sources Assistant can access and set permissions under Administration > Grafana Assistant.
Step 5: Review Conversation History and Tool Usage
Assistant transparently displays the full conversation and any tool calls it made. Errors or warnings from those tools are fed back into the chat, allowing you to debug or iterate. This feature helps you trust its outputs and learn from its interactions. Use it to refine future queries.
Tips for Getting the Most Out of Grafana Assistant
- Start simple: Begin with basic questions about your dashboards or alerts before diving into complex incident investigations.
- Provide context: Mention which data source or time range you're interested in—Assistant works best with clear specifications.
- Leverage conversation history: Review the “show its work” output to understand how Assistant arrived at an answer; this helps you build trust and improve its responses.
- Use natural language variations: If one phrasing doesn't work, rephrase—Assistant understands multiple ways of asking for the same thing.
- Monitor usage limits: The free plan has generous but finite limits. Check your Grafana Cloud dashboard to avoid surprises.
- Stay updated: Grafana often releases improvements to Assistant; keep your plugin and Grafana version current to benefit from new features.