Evaluate Self-Hosted

Spin up a self-hosted Pipekit deployment on a dev cluster in 30 minutes.

A 30-minute path to a working Self-Hosted Pipekit on a local Kubernetes cluster. The goal is to let you click around the UI, run a sample workflow, and decide if Self-Hosted is the right shape for you. It does not set up a production install.

For a production deploy, follow Deploy to Production instead.

Prerequisites

• kubectl, helm (v3), and docker installed locally.

• A local Kubernetes cluster (k3d, kind, minikube, or Docker Desktop Kubernetes all work). This guide uses k3d.

• A Pipekit license key for the trial. Email [e-mail] if you don't have one; we'll send you a 30-day trial key.

• Access to the pipekitprivate container registry on Docker Hub. Email [e-mail] if you don't have access; we'll set it up.

1. Spin up a local cluster

k3d cluster create pipekit-eval \
  --port "30000:30000@loadbalancer" \
  --port "30001:30001@loadbalancer"

This gives you a single-node cluster reachable via kubectl. For other local cluster tools, the install steps are identical. Only the cluster-creation command changes.

2. Install Argo Workflows

Pipekit runs on top of Argo Workflows; install it first.

kubectl create namespace argo
kubectl apply -n argo -f https://github.com/argoproj/argo-workflows/releases/latest/download/quick-start-minimal.yaml

Wait for the Argo pods to be Ready:

kubectl get pods -n argo -w

3. Pull Pipekit container images

The Self-Hosted install uses private images from the pipekitprivate Docker Hub registry.

docker login -u <your-docker-username>
kubectl create namespace pipekit
kubectl create secret docker-registry regcred \
  --docker-username=<your-docker-username> \
  --docker-password=<your-docker-password> \
  --namespace=pipekit

For the full image list, see Container Images.

4. Install the Pipekit Helm chart

helm repo add pipekit https://helm.pipekit.io
helm repo update

helm install pipekit pipekit/pipekit \
  --namespace=pipekit \
  --set global.deployType=dev \
  --set license.key="<your-trial-license-key>" \
  --set imagePullSecrets[0].name=regcred

This deploys the Pipekit UI, API, Pipekit Agent, ID service, and Postgres into the pipekit namespace.

For the full values reference, see Reference > Helm Chart Values > Self-Hosted Chart.

5. Wait for everything to be Ready

kubectl get pods -n pipekit -w

This takes 3 to 5 minutes on a fresh cluster. Postgres and the Pipekit services need to come up in order.

6. Log in with the break-glass account

The first install creates a single Org Admin account with a randomly generated password printed to the install logs. You can also fetch it from the cluster:

kubectl get secret -n pipekit pipekit-break-glass -o jsonpath='{.data.password}' | base64 -d
echo

Browse to http://localhost:30000 and sign in as [e-mail] with that password.

For the full break-glass setup, see Configure > Initial Login & Break-Glass.

7. Submit a sample workflow

In a separate terminal:

brew install pipekit/tap/cli
pipekit login --host=http://localhost:30000
git clone https://github.com/pipekit/examples.git
cd examples
pipekit submit --cluster-name=default examples/dag-diamond/workflow.yaml

Open the Run in the UI and confirm logs and the Run graph render correctly.

What's next

• Production install: replace deployType=dev with prod, set up real ingress, point at a managed Postgres, configure backups. Follow Deploy to Production.

• Configure SSO: wire your identity provider in. See Configure > SSO.

• Connect a Git provider: let users trigger Workflows from commits and pull requests. See Configure > Git Provider.

• Right-size resources: the dev profile is too small for real workloads. Bump replica counts, resource requests, and HPA limits via Helm values.

Tearing down

helm uninstall pipekit -n pipekit
kubectl delete namespace pipekit
k3d cluster delete pipekit-eval