About K8sGPT
K8sGPT is an open source tool that scans a Kubernetes cluster, finds what is broken and explains the problem in plain English with suggested next steps. It is a CNCF Sandbox project released under the Apache 2.0 license, and it runs from three surfaces: a command line tool, an in cluster operator and an MCP server for use inside AI assistants.
The core engine ships with 23 built in analyzers that encode site reliability engineering knowledge and run without any AI at all. Adding the explain option sends findings to a language model backend, and the vendor lists 14 supported backends, both cloud hosted and local. An anonymize option masks cluster data before it is sent. Custom analyzers can be written in any language over gRPC, and findings from tools such as Trivy can be brought in as well.
The operator is configured with a single custom resource. It watches the cluster on an interval, diagnoses issues with the chosen backend, records findings as Result resources, routes notifications to sinks such as Slack and, when enabled, performs guarded auto remediation governed by a risk threshold. The CLI installs through Homebrew and is pointed at a kubeconfig, then authenticated against a chosen AI backend.
The software is free and open source. Any cost of the chosen AI backend, such as an API key for a hosted model, is separate from K8sGPT itself. The vendor does not publish paid plans.