New Relic has launched the Autopilot Kubernetes Agent, an AI-powered tool that diagnoses faults in Kubernetes clusters and drafts fix code for engineers, and it is available now in public preview. The company unveiled the agent at KubeCon North America 2026 in Salt Lake City, Utah, where it is being demonstrated to operations teams from 9 to 12 November.

The agent is aimed at site reliability engineers and platform engineers who run containerised workloads and need to find the cause of an incident quickly. Kubernetes environments can be hard to troubleshoot because a single failure may involve compute resources, scheduling, networking or cloud configuration, and diagnosing it often requires deep specialist knowledge of how clusters behave.

What the Kubernetes agent does

The Autopilot Kubernetes Agent automatically pinpoints the root cause of a cluster incident and generates context-aware fix code, which engineers review and then run themselves. It is a specialised extension of New Relic Autopilot, the company's wider fleet of AI agents that investigate incidents and speed up response workflows.

According to New Relic, the agent runs more than 90 specialised diagnostic detectors across Kubernetes events, object descriptions, logs and application performance monitoring traces. By correlating those signals, it isolates the likely cause of a problem and produces ready-to-run remediation code rather than leaving engineers to piece together clues from separate dashboards.

The agent is powered by New Relic's Ground Truth context engine, which evaluates operational data built on OpenTelemetry, the open standard for collecting telemetry from software systems. New Relic says this grounding in real operational data is what lets the agent produce fixes matched to the specific cluster it is examining.

Why Kubernetes troubleshooting needs help

Kubernetes has become the standard platform for running modern applications, and the pressure on the engineers who operate it is growing. The Cloud Native Computing Foundation reports that 66 percent of organisations running generative AI rely on Kubernetes, which means clusters now carry some of the most demanding and business-critical workloads in use.

At the same time, the skills needed to run those clusters are in short supply. Incident detection and resolution in complex container environments traditionally depend on engineers who understand cluster behaviour in depth, and when those people are unavailable, outages can last longer and cost more. New Relic describes the new agent as a way to close that expertise gap, giving less specialised engineers a starting point for diagnosis and repair.

Public preview and availability

The New Relic Autopilot Kubernetes Agent is available in public preview, and attendees at KubeCon North America 2026 can see it working live at the company's stand in Salt Lake City. New Relic's chief product officer, Brian Emerson, said the agent embeds specialised agentic AI directly into engineers' containerised workflows, removing guesswork from complex cluster incidents while keeping the engineer in control of any fix.

Because the tool is still in preview, teams should treat its suggested fixes as a starting point and check them carefully before applying them to production systems. New Relic has not published independent measurements of how accurately the agent's detectors identify root causes in live clusters, so its performance in real incidents remains to be proven outside the company's own demonstrations.

The launch comes as observability vendors race to add AI agents to their platforms, aiming to move beyond dashboards and alerts towards tools that actively investigate problems. For teams already running New Relic, the Kubernetes agent arrives as an extension of the Autopilot system they may already use; for others, the public preview is a chance to test whether AI-generated remediation can shorten the time between an alert and a fix.