Grafana Labs, the company behind the open observability cloud, announced a set of new AI-focused capabilities at GrafanaCON 2026: AI Observability in Grafana Cloud; a significant expansion of Grafana Assistant into more environments, as well as new agentic capabilities; the Grafana Cloud CLI (GCX), a new agentic interface for automated and agent-driven workflows; and o11y-bench, a new open source benchmark for evaluating AI agents running observability workflows.AI is transitioning from experimentation to production, while observability, control, and operational trust are still catching up. Grafana Labs’ 2026 Observability Survey found near-universal interest in AI’s value, alongside real caution about autonomy: 15% of respondents expressed skepticism about AI taking autonomous actions without stronger safeguards. Now, observing LLMs and the systems they touch is becoming table stakes for teams that intend to run them safely and reliably at scale.“AI systems are starting to look a lot like distributed systems did a decade ago: powerful, but difficult to reason about and even harder to operate,” said Jen Villa, senior director of Product, Grafana Labs. “We’re not approaching this as a separate category. The goal is to bring the same level of visibility and control to AI that teams already expect from the rest of their stack.”

Introducing AI Observability in Grafana Cloud: Monitor and Evaluate AI Systems in Real Time
Launching in Public Preview, AI Observability in Grafana Cloud is a complete solution designed to help teams monitor and evaluate LLM-powered applications and agents in real time.

As AI becomes embedded in customer-facing experiences, failures often don’t look like classic telemetry: unexpected outputs, inconsistent behavior, and silent degradation that erodes trust before traditional dashboards light up.

AI Observability in Grafana Cloud is built to close these visibility gaps by helping teams:

• Observe AI agent behavior in real time, including inputs, outputs, and execution flows
• Continuously evaluate outputs, with alerts for issues such as low-quality responses, policy violations, or anomalous behavior
• Surface risk earlier, including potential data exposure or misuse (for example, leaked credentials or abnormal usage patterns)
• Elevate agent sessions and conversations to first-class telemetry signals and correlate them in the same environment where applications are observed

Teams can get started with AI Observability in Grafana Cloud today to understand what their AI is doing, how well it’s doing it, and where issues are emerging.

Grafana Assistant: Broader Reach, Deeper Workflow Support
Grafana Labs also announced a significant expansion of Grafana Assistant, its AI-powered agent for observability and operational workflows that helps monitor, troubleshoot, and manage systems through natural language conversations.

Assistant is no longer limited to Grafana Cloud; it is extending to additional environments and surfaces, including on-premises deployments of Grafana Enterprise, so teams with stricter data and control requirements can use the same AI-assisted workflows. Grafana open source users will also be able to use Grafana Assistant by connecting their accounts to a Grafana Cloud instance.

Additional new capabilities coming to Grafana Assistant include:

• Assistant Workspace: Bring Grafana Assistant into full-screen, chat, and browse visualizations at the same time
• Assistant API: Call Grafana Assistant in workflows from anywhere and move the Assistant into your stack
• Automations: Schedule tasks and automation workflows, enabling routine operational actions to run without manual intervention
• Remote MCP server: Bring any agent and connect it to Grafana’s remote MCP server
• Learn mode: Get personalized, hands-on lessons tailored to your role and infrastructure so you can elevate your skills
• And so much more: Grafana Assistant in Microsoft Teams, 50+ integrations, 15 native data source integrations, Python runtime in the Assistant, and EU-preferred inference for European customers

The focus is shortening the distance from question to grounded investigation, especially when minutes matter. It is not just a generic chat interface. 

To get started with Grafana Assistant, Grafana Enterprise and Grafana OSS users can create a Grafana Cloud account (including the actually useful free forever plan) and connect it to their Grafana installation via a one-click setup.

Grafana Cloud CLI (GCX): Observability Where Agents Already Work
Grafana Labs also introduced the Grafana Cloud CLI (GCX), a new interface designed for a shift already underway in how software is built and operated: engineers are increasingly working through AI-assisted development environments like Cursor, Claude Code, and GitHub Copilot, where the agent becomes the primary interface.

GCX is designed to bring Grafana Cloud into that workflow, so teams can:

• Access the full Grafana Cloud surface area agentically, including provisioning, configuration, and telemetry querying
• Invoke Grafana Assistant capabilities from the dev environment, without context-switching into separate tools
• Close the agentic loop between code and production by using agents to query live observability insights, correlate alerts with recent repository changes, and propose fixes without leaving the development environment to create a continuous feedback cycle where observability data drives the next action

The intent is fewer handoffs between code, alerts, and dashboards, so investigation and remediation can stay closer to where changes ship. Download GCX and get started today. 

o11y-bench: An Open Benchmark for Observability Agents
Grafana Labs also announced it is open sourcing o11y-bench, a benchmark for evaluating AI agents on observability workflows.

Built on Harbor and designed to run vs. a real Grafana stack, o11y-bench is intended to help teams measure how agents perform on the kinds of tasks that matter in practice: querying metrics, logs, and traces; investigating incidents; and making targeted dashboard changes.

In modern observability environments, where teams operate across open tools and multiple telemetry types, evaluating AI agents requires more than reviewing outputs. o11y-bench is designed to reflect that reality by measuring what agents actually do in the system, not just what they say.

Learn more at https://o11ybench.ai/ 

Building the Foundation for AI in Production
Together, these moves reflect a pattern: AI in production adds complexity that traditional tooling only partly covers. To coordinate work across AI observability, assistant experiences, and agent-driven workflows, Grafana Labs is forming a dedicated AI organization unifying these efforts under one team.

Mat Ryer has been appointed Senior Director of AI and will lead this work across the company.

“AI breaks in ways traditional observability wasn’t designed for,” said Mat Ryer, senior director of AI, Grafana Labs. “Latency and errors still matter, but they’re not enough. You also need visibility into correctness, consistency, and shifting agentic behaviours over time. AI is quickly becoming a key part of the way teams investigate and operate systems. We want to make all of that observable in a way that’s practical, reliable, and fits into how engineers already work today.”