Grafana Cloud

Use Assistant Search

Note

Assistant Search is currently in public preview. Grafana Labs offers limited support, and breaking changes might occur prior to the feature being made generally available.

Assistant Search helps you find indexed Grafana resources and Assistant memory by describing what you need. Instead of remembering exact names, folders, or labels, you can search in natural language across supported Grafana content.

What you’ll achieve

Use Assistant Search when you want to quickly find resources related to a service, symptom, incident, or operational area. It can return matching dashboards, incidents, investigations, infrastructure context, and alert rules when those collections are available for your stack.

For example, you can search for “payment service latency” to find related dashboards, indexed Grafana IRM incidents, previous Assistant investigations, and alert rules that match the same area.

Before you begin

Before you use Assistant Search, make sure Grafana Assistant is available for your Grafana Cloud stack.

Assistant Search follows Grafana access controls. Results are filtered by the resources you can access, including dashboard and data source permissions where applicable. If you can’t access a resource directly in Grafana, Assistant Search doesn’t use it to answer your query.

Search Grafana resources

Navigate to AI > Assistant in the main menu and describe the resource, system, or problem you want to find in chat. For example, ask “Find dashboards for checkout errors.”

Assistant searches the relevant collections and returns results in your conversation. You can also start a search from a chat in Workspace.

Assistant Search looks across supported collections and returns the closest matches. Depending on your stack configuration, results can include:

  • Dashboards and dashboard panel context.
  • Grafana IRM incidents.
  • Investigations.
  • Infrastructure context from Assistant memory.
  • Alert rules.

Search results link back to the source resource when a direct link is available.

Choose how many results to retrieve

In chat, you can request a different result limit for each collection, for example, “Find up to five dashboards and three incidents related to checkout errors.”

You can request between 1 and 200 results per collection. If you don’t specify limits, search retrieves up to 20 dashboard results, three infrastructure results, five investigations, five incidents, and 20 alert rules. Access checks, relevance filtering, and duplicate removal can return fewer matches. These limits don’t guarantee that many distinct resources.

Search incident and investigation history

Incident and investigation searches default to records created in the last three calendar months. To search older records or a specific period, include the time range in your request, for example, “Find checkout investigations created between January and March 2026.”

The time range filters when the incident or investigation record is created, not when the underlying problem occurs. It doesn’t restrict dashboard, infrastructure, or alert rule results.

Investigation results include a summary of the findings when available. If there are no findings, the result uses the investigation’s objective instead. Open the linked investigation to review the full context.

Example searches

Use natural language to search for operational areas, symptoms, services, or resources. For example:

payment service latency

dashboards for checkout errors

previous investigations about Mimir query latency

alert rules for auth service saturation

infrastructure context for query frontend memory

incidents related to Kafka timeouts

Continue from search results

Ask follow-up questions in the same conversation to explore a result or investigate a related issue. Review the linked sources before you make operational decisions.

Next steps