en.Wedoany.com Reported - Grafana Labs released a range of new artificial intelligence capabilities during its AI Week event held from July 27 to 31, including new ways to interact with telemetry data, investigate incidents, and monitor the behavior of AI agents in production environments. Mat Ryer, the company's Senior Director of AI, stated that the release aims to showcase tools that have already proven useful to customers, rather than features built to capitalize on market interest in AI.
Ryer noted that the company was a late entrant in the AI space, which allowed it to observe the effectiveness of existing solutions in the market. He mentioned that many current products contain speculative elements, and the company is not interested in AI hype, ensuring that it delivers concrete, real capabilities that can make a tangible difference for customers before release.
Earlier this week, Grafana Labs announced the general availability of its Workspace tool. This tool provides an interface built around conversational agents, replacing manually constructed dashboards. Users start with a conversation, ask questions in natural language, and the agent responds based on telemetry data.
The company also officially released the command-line tool gcx along with cloud and open-source MCP servers, which previously existed only in early versions with limited functionality. These tools bring Grafana's agent capabilities to programming agents, enabling planned code changes to be based on real production telemetry data rather than being developed in isolation from telemetry. General availability means the underlying APIs are now finalized and come with backward compatibility commitments that were previously absent.
Grafana Labs' Assistant Investigations feature became generally available alongside Assistant Automations, while a third capability, Watcher agents, moved to public preview. The Investigations feature deploys multiple specialized agents to examine alerts by tracking leads and testing possible explanations, which are then reviewed by humans. The process generates a series of hypotheses and attempts to disprove them, incorporating an adversarial element where agents inherently question themselves to increase confidence in their conclusions.
The Automations feature allows teams to schedule natural language reports, such as weekly summaries of incident themes, delivered directly to Slack or Microsoft Teams. Watcher agents continuously monitor situations that are difficult to define with fixed metrics or thresholds, and will take autonomous action after users test and confirm their behavior.
On Thursday, Grafana Labs announced the general availability of its Agent Observability capability. This feature gives teams insight into the latency and cost of AI agents, as well as the substance of their conversations, including whether they provide sound advice and whether they use the correct tools. The following day, the company will extend Grafana Assistant to Slack, Microsoft Teams, and mobile and desktop applications.
AI Week was scheduled to fill the gap between Grafana Labs' two annual conferences—GrafanaCon and ObservabilityCON, held in San Francisco in October. Ryer noted that most customer engagement surrounding these releases occurs in community channels such as Reddit and podcast recordings, and the company encourages customers to test tools with their own production data rather than relying on demonstrations. The overall goal is to enable teams to verify that agent outputs are trustworthy rather than accepting them blindly. Ryer stated that this helps users genuinely believe that agents are performing their intended tasks and that their results can be relied upon.
