k8m

Why we included this project

Most Kubernetes pain shows up as context juggling and terminal sprawl, and k8m is a dashboard built to end both: one console manages several clusters, while an embedded large language model explains resources, interprets Describe output, diagnoses logs, and suggests commands in natural language. The MCP layer is the part that stands out. It ships cluster-aware tools so the model can actually perform operations, and permissions follow the logged-in user, so an agent never acts beyond its authority. The same tool handles routine chores: scheduled inspections with custom Lua rules, event forwarding to chat or webhooks, CRD browsing, Helm installs, and pod file and log access. SREs and platform engineers who want a self-hosted dashboard with real AI assistance, not a demo, will find one that deploys as a single cross-platform binary they can spin up inside a cluster or on a workstation.