Quick Start Guide

Prerequisites

• A Kubernetes cluster and kubectl access with permission to install CRDs
• An AI provider API key (e.g. Anthropic — see AI Provider Integration)
• A GitHub repository you can push branches to
• A Nominos account (early access)

Step 1: Install the Operator

nominos-core is the operator that runs the agents. Your Nominos onboarding contact will provide CRD and controller install instructions for your cluster.

Verify the CRDs and controller are up:

kubectl get crds | grep nominos.io
# agentjobs.nominos.io, agentplans.nominos.io, mergequeues.nominos.io, ...

kubectl get pods -n nominos-system

Step 2: Connect Your Cluster

In the Nominos dashboard, go to Settings → Connections → Kubernetes, click Add Cluster, and upload or paste your kubeconfig. Use Test Connection to confirm. Full details in Kubernetes Setup.

Step 3: Add Your Provider API Key

Go to Settings → Connections → AI Providers, select your provider, and paste your API key. Agents run with your key on your cluster — Nominos never proxies or marks up your API usage. See AI Provider Integration.

Step 4: Create Your First Job

Create a job from the dashboard (Jobs → New Job), or apply an AgentJob directly:

apiVersion: nominos.io/v1
kind: AgentJob
metadata:
  name: first-job
spec:
  task:
    description: "Add a CONTRIBUTING.md with setup and test instructions"
  repository:
    url: https://github.com/your-org/your-repo
    createPR: true

kubectl apply -f first-job.yaml

Prefer the CLI? nomctl can create the same job from inside any git checkout:

# One-time setup
nomctl login

# Create a job for the current repo and wait for it to finish
nomctl run "Add a CONTRIBUTING.md with setup and test instructions" --pr --wait

Step 5: Watch It Open a PR

Track the job as it moves through its phases (Pending → Queued → Running → Completed):

# Live phase updates
kubectl get agentjobs -w

# Stream agent logs
nomctl job logs first-job --follow

# Where's my PR?
kubectl get agentjob first-job -o jsonpath='{.status.pullRequestURL}'

The same run is visible in the dashboard's job detail view with live logs, cost, and token usage. When the job completes, status.pullRequestURL links to a pull request on your repo — review and merge it like any other PR, or route it through a MergeQueue.