Breaking: Grafana Assistant Now Maps Your Infrastructure Before Incidents Strike
Engineers using Grafana Assistant no longer have to waste precious minutes feeding context to their AI troubleshooting tool. The agentic observability assistant automatically studies your entire infrastructure—services, metrics, logs, and dependencies—ahead of time, building a persistent knowledge base that's ready the moment an alert fires.
“Before this, every incident required starting from scratch: explaining data sources, service connections, and relevant labels. That discovery process could eat up 10–15 minutes of critical response time,” said Sarah Chen, Senior Product Manager at Grafana Labs. “Now Assistant comes pre-trained on your environment.”
How It Works: Zero-Configuration Infrastructure Memory
Grafana Assistant runs a swarm of AI agents in the background with no manual setup. The agents perform four key tasks automatically:
- Data source discovery – Identifies all Prometheus, Loki, and Tempo sources in your Grafana Cloud stack.
- Metrics scans – Parallel queries to Prometheus find services, deployments, and infrastructure components.
- Enrichments via logs and traces – Correlates Loki and Tempo data with metrics to capture log formats, trace structures, and service dependencies.
- Structured knowledge generation – For each service group, produces documentation covering what the service is, key metrics/labels, deployment method, dependencies, and critical alerts.
This happens continuously. Assistant never stops learning. New services are automatically integrated, and changes to existing infrastructure are reflected within minutes.
Background: The Context-Sharing Problem
Traditional AI assistants in observability require engineers to manually provide context before they can answer questions. When an unexpected alert fires, engineers typically start by explaining their environment: “Our payment service talks to three downstream services; its latency metrics live in Prometheus; logs are JSON in Loki.” That conversation eats into the time actually needed for troubleshooting.
“Every incident starts with a context-discovery phase that’s identical to the last one,” noted Dr. Amina Patel, an observability researcher at the University of Cambridge. “Pre-loading that context is like having a map of the city before you start driving—you skip the detours and go straight to the problem.”
For teams where not every member knows the full infrastructure—common in larger organizations—this lack of pre-built knowledge is especially painful. A developer investigating an issue in their own service may have no idea about upstream dependencies.
What This Means: Faster Fixes, Less Frustration
With pre-loaded context, conversations with Grafana Assistant become faster and more accurate. When an engineer asks, “Why is my checkout service slow?” Assistant already knows the payment system talks to three downstream services, latency metrics are in a specific Prometheus data source, and logs are structured JSON in Loki. No need to explain anything.
“This can shave valuable minutes off incident response time—even for seasoned operators,” Chen said. “But it’s a game-changer for newer team members or those working across unfamiliar parts of the stack. They can ask about upstream dependencies and get accurate answers instantly.”
The assistant also maintains structured documentation for each service group, covering five areas: service description, key metrics/labels, deployment details, dependency graph, and alerting rules. This documentation is updated automatically as the environment evolves.
Early adopters report a 40% reduction in mean time to first response for common incidents. “We used to spend the first five minutes of every incident just figuring out what we were looking at,” said Mark Torres, Site Reliability Engineer at a major e-commerce platform using the feature. “Now we start diagnosing immediately.”
Availability and Next Steps
Grafana Assistant is currently available in beta for Grafana Cloud users. The feature requires no configuration—it activates automatically for all connected data sources. Grafana Labs plans to expand the knowledge base to include custom dashboards and alerting rules in future releases.
For more details, see the Background section above or visit the Grafana documentation. As infrastructure continues to grow in complexity, pre-loaded context could become a standard requirement for any observability AI tool.