1. What is Grafana Assistant?
Introduction to artificial intelligence in Grafana Cloud.
A copilot for observability
Grafana Assistant is a tool powered by artificial intelligence (using LLMs) and natively integrated into Grafana Cloud. As part of Grafana's Machine Learning suite, it acts as a copilot for DevOps and SRE engineers, allowing them to interact with their telemetry data using natural language. It aims to drastically reduce the learning curve for querying languages (PromQL, LogQL, etc.) and accelerate mean time to resolution (MTTR).
2. Main features
Generation, query explanation, and contextual analysis.
What the Assistant can do for you
One of the major strengths of the Assistant is its ability to understand and generate complex queries:
- Query generation: Ask in natural language (e.g., "Show me the HTTP 500 error rate per service over the last 3 hours") and the AI generates the corresponding PromQL, LogQL, or TraceQL query.
- Code explanation: Did you inherit an old dashboard with an unreadable query? The Assistant can break it down and explain it step by step.
- Incident analysis: The Assistant can summarize the state of a dashboard or help investigate an alert by suggesting the logs or traces associated with the ongoing issue.
3. How to use the Assistant properly (Prompting)
Adopting the right approaches to get relevant results.
Optimizing interactions with AI
To get the most out of Grafana Assistant, it's important to understand how to interact with the AI model and guide it with the right context. Here are some tips to optimize your prompts:
- Use @ mentions: Precisely target your data sources (e.g., @Prometheus, @Loki, @Tempo) to give immediate context to the AI.
- Be specific: Explicitly mention the names of metrics, desired labels (e.g.,
env="production"), or aggregation types to improve query efficiency. - Rephrase if needed: The AI is iterative. If the first answer isn't perfect, refine your query by adding details or rephrasing it.
Best Practice: The richer and more precise the context, the more exact the generated query will be on the first try. Combining @ mentions and precise metric names is the key.
Common Mistake: Blindly executing an AI-generated query on a highly loaded environment without verifying it. A poorly filtered LogQL query generated by the Assistant can be very resource-intensive and slow down your queries. Always verify the time filters and the presence of indexed labels before execution.
4. Practical part: Access and interact with the Assistant
How to use Grafana Assistant from the interface, dashboards, and the explorer.
Accessing the Assistant
Grafana Assistant is accessible in several strategic places within the interface: via the main menu (often represented by a sparkles icon ✨), directly in the Explore interface for query creation assistance, or from your Dashboards.
Integration from a Dashboard
On any panel of your dashboard, you can invoke the Assistant to interact with the displayed data. Click on the panel menu (the three dots) and select the Assistant. You can then ask it to explain the underlying query, suggest optimizations, or analyze any anomaly detected over the displayed time period.
Using in Explore
In the Explore tab, you can use the Assistant's text box to formulate your needs in natural language. Instead of writing complex LogQL, just type: "Show me the errors on the frontend pod", and the Assistant will pre-fill the query box. You can then iterate and refine the query before executing it.
5. Security and Privacy
How your data is processed by the AI model.
Protection of telemetry data
The use of Artificial Intelligence legitimately raises privacy questions, particularly for corporate data. Grafana Labs is committed to security: your dashboard data, the results of your queries, and your logs are not used to train base models (LLMs) of third-party entities. The Assistant relies on the context you provide in an ephemeral and secure manner, strictly isolated within your Grafana Cloud instance.
6. Availability
Grafana Assistant is available for all editions.
Opened to OSS and Enterprise
Since the announcements at GrafanaCON 2026, Grafana Assistant is no longer limited to Grafana Cloud. The tool is now globally available for all editions: Grafana Cloud, Grafana OSS (Open Source), and Grafana Enterprise. This allows all teams to benefit from the power of generative AI directly in their local (on-premise) or hybrid environment to drive their observability in natural language.