Control Assistant data sources

Grafana Assistant bases every answer on the data sources that already power your Grafana Cloud dashboards. This article explains how the Assistant connects to those data sources, which telemetry types it understands, and how discovery features keep conversations grounded in the sources you trust.

Before you begin

Grafana Cloud stack: Enable Grafana Assistant in a Grafana Cloud instance that already collects the telemetry you want to analyze.
Data source permissions: Make sure you can query the data sources directly in Grafana. The Assistant inherits the same access controls, so it can only reach sources you can reach.
Consistent naming: Keep data source titles, folders, and tags descriptive so discovery workflows return the sources you expect.

How Grafana Assistant connects to data sources

The Assistant never bypasses Grafana. When you ask a question, Grafana routes the request through the same data source proxy endpoints that dashboards and Explore use. The backend records each call as a usage event with the model, token counts, and estimated cost before forwarding the response to the chat. Because the Assistant relies on Grafana Cloud APIs, you get audit logging, encryption, and network controls identical to the rest of your stack.

Supported telemetry types

The Assistant can work with any data source that Grafana exposes through the supported tools. Key categories include:

Metrics: Prometheus data sources, including managed Grafana Cloud Prometheus instances and on-premises endpoints made available through Cloud connections.
Logs: Loki data sources, so you can pivot from metric spikes to log samples without leaving the chat.
Traces: Tempo data sources for latency breakdowns, trace comparisons, and span-level filtering.
Profiles: Pyroscope data sources supplied through Grafana Cloud Profiles, enabling questions about CPU, memory, and other profiling data.
Load testing: k6 Cloud results for post-test summaries and trend analysis.
SQL: Managed and self-hosted SQL data sources that use Grafana’s SQL APIs, including PostgreSQL, MySQL, ClickHouse, and BigQuery connectors.

The Assistant also respects custom data sources surfaced through investigations or discovery scans, as long as those data sources expose metadata Grafana can consume.

Source selection during chat

Grafana Assistant detects relevant data sources automatically based on your question, prior chat context, and any memories loaded into the conversation. You can steer the response with @ mentions to pick exact data sources, dashboards, or panels. When a conversation pivots to a different system or environment, start a new chat so the Assistant chooses data sources that match the new topic.

Discovery and context

Discovery workflows enumerate data sources, metrics, labels, and infrastructure relationships so the Assistant can preload accurate context. Run a discovery scan when you connect a new environment to Grafana Cloud, then apply the saved memories during future chats to keep responses aligned with the latest topology. For detailed steps, refer to Use infrastructure memory for context.

Governance and best practices

Grafana Cloud enforces data source access end to end: the Assistant inherits your user permissions, and requests stay inside Grafana’s audited API surface. Grafana Labs captures token counts, tool execution details, and optional cost estimates without storing raw query results outside Grafana Cloud. Combine descriptive data source metadata with access reviews to keep Assistant conversations scoped to the right systems.

Next steps

Understand RBAC access control so you can manage who uses the Assistant.
Review Manage your data privacy and security to learn how data handling, third-party providers, security measures, and access controls work.