Quick Start

Install the CLI. Point it at Kubernetes. Run an agent.

The fastest path into witwave starts with ww, the command-line tool that installs the operator, creates agents, sends prompts, and shows logs and events without needing a browser or local port-forward.

Before you start

You need a Kubernetes cluster.

Docker Desktop, Rancher Desktop, kind, minikube, or a remote cluster can work. The important part is that your current kubeconfig context points at a cluster where you can install CRDs, create namespaces, and run pods.

01

Install

Install ww.

The universal installer works on Linux and macOS and verifies the release checksum.

bash
curl -fsSL https://github.com/witwave-ai/witwave/releases/latest/download/install.sh | sh

macOS alternative:

bash
brew install witwave-ai/homebrew-ww/ww
02

Confirm

Check your CLI and cluster context.

Make sure ww is on your path and your kubeconfig is pointed at the cluster you intend to use.

bash
ww version
kubectl config current-context
03

Operator

Install the witwave operator.

The operator reconciles WitwaveAgent, WitwavePrompt, and WitwaveWorkspace resources into running infrastructure.

bash
ww operator install --if-missing --yes
ww operator status
04

Hello agent

Create and ping an echo-backed agent.

The default echo backend needs no LLM API key, so it is the safest first smoke test before wiring Claude, Codex, Gemini, workspaces, or git-backed configuration.

bash
ww agent create hello --namespace witwave --create-namespace
ww agent status hello --namespace witwave
ww agent send hello "ping from quick start" --namespace witwave

Next

Go from hello-world to a real agent.

The walkthrough continues into multi-backend agents, credentials, GitOps wiring, logs, events, and cleanup. Use this page to get moving; use the docs when you are ready to shape an actual agent team.

Read the CLI reference