Quick Start Guide

Get from an empty cluster to a working AI debug chat in about a minute.

Prerequisites

  • A Kubernetes cluster (Kind, Minikube, EKS, GKE, etc.) with kubectl pointed at it
  • helm 3.x and Node.js 18+ installed
  • An LLM API key (Anthropic, OpenAI, or Google)

The one-command path

npm install -g @kubently/cli
kubently install

That’s it. The installer:

  1. Creates the kubently namespace and all secrets (API keys, Redis password, LLM key)
  2. Installs the Helm chart from the published repo (API + Redis + executor)
  3. Waits for the executor to connect, port-forwards the API
  4. Saves your CLI config and drops you straight into a debug chat

It reads your LLM key from ANTHROPIC_API_KEY / OPENAI_API_KEY / GOOGLE_API_KEY, or prompts for it. Useful flags:

kubently install --provider openai        # pick your LLM provider
kubently install --yes --no-chat          # non-interactive (CI, scripts)
kubently install --cluster-id prod-east   # name the cluster yourself
kubently install --help                   # everything else

Start debugging

The installer ends in a chat session. To come back later:

kubectl -n kubently port-forward svc/kubently-api 8080:8080 &
kubently debug

Ask natural language questions like:

“What pods are running in the kube-system namespace?” “Are there any pods in a CrashLoopBackOff state?” “Check the logs for the api-server pod.”

Use it from Claude Code or Cursor

Kubently is also an MCP server — add it to your AI editor in one line:

claude mcp add kubently -- kubently mcp

See the MCP guide for Cursor and generic client config.

Prefer to do it by hand?

The installer is just automation over the standard Helm flow. For manual installation, production values, TLS, and multi-cluster executor setup, see the Installation Guide.

Next Steps

  1. Read the Installation Guide — production deployment details.
  2. Read the CLI Guide — complete CLI documentation.
  3. Explore the API Reference — build custom integrations.