Deploy CAIPE with Helm

This guide helps you deploy CAIPE (Community AI Platform Engineering) on any Kubernetes cluster using Helm. No prior experience with CAIPE is required.

What is CAIPE? CAIPE is an open-source platform for building and running AI agents that use tools, LLMs (e.g. Claude or GPT), and multi-agent orchestration. The Helm chart deploys the supervisor, UI, and optional agents (ArgoCD, GitHub, Backstage, RAG, etc.) on your cluster.

When to use Helm: Use this path when you already have a Kubernetes cluster (EKS, GKE, AKS, KinD, etc.) and want to install CAIPE from the official chart. For a one-command local setup, see Run CAIPE with KinD instead.

Step 1: Clone the repository (optional but recommended)

Cloning the repo gives you the chart source, examples, and EKS/config references:

git clone https://github.com/cnoe-io/ai-platform-engineering.git
cd ai-platform-engineering

You can install the chart directly from the OCI registry (no clone required). Cloning is useful for customising values or using a values file from the repo.

Step 2: Prerequisites

Before installing the chart, ensure you have:

Requirement	Purpose
Kubernetes cluster	Version 1.28 or higher (EKS, GKE, AKS, KinD, etc.)
kubectl	Configured to access your cluster (`kubectl get nodes` should work)
Helm 3.x	To install and upgrade the chart
Credentials	API keys and secrets for the agents you enable (see Configure Agent Secrets)

You must configure secrets before or right after installing the chart so that agents can authenticate to external services. See Configure Agent Secrets for details.

Step 3: Install the chart

The chart is published to GitHub Container Registry (GHCR). You can install without cloning the repo.

Chart version in this guide: 0.2.32. For the latest chart version, see GitHub Releases—then replace 0.2.32 with the release tag (e.g. 0.2.33) in the commands below.

Basic installation (ArgoCD, Backstage, GitHub agents)

helm install ai-platform-engineering oci://ghcr.io/cnoe-io/charts/ai-platform-engineering \
  --version 0.2.32 \
  --namespace ai-platform-engineering \
  --create-namespace \
  --set-string tags.basic=true

Complete profile (all agents)

helm install ai-platform-engineering oci://ghcr.io/cnoe-io/charts/ai-platform-engineering \
  --version 0.2.32 \
  --namespace ai-platform-engineering \
  --create-namespace \
  --set-string tags.complete=true

After installation, configure your agent secrets and LLM provider if you haven’t already. Then use kubectl get pods -n ai-platform-engineering to confirm pods are running.

Step 4: Verify the deployment

# List Helm releases in the namespace
helm list -n ai-platform-engineering

# Check pod status
kubectl get pods -n ai-platform-engineering

# View logs for a specific agent (example: GitHub agent)
kubectl logs -n ai-platform-engineering -l app=agent-github

Customising the deployment

The chart uses tags to enable or disable components. Two built-in profiles:

Profile	Tag	What’s included
Basic	`tags.basic=true`	ArgoCD, Backstage, GitHub agents
Complete	`tags.complete=true`	All agents and RAG stack

Add specific agents

Combine the basic profile with extra agents:

helm install ai-platform-engineering oci://ghcr.io/cnoe-io/charts/ai-platform-engineering \
  --version 0.2.32 \
  --namespace ai-platform-engineering \
  --create-namespace \
  --set-string tags.basic=true \
  --set-string tags.agent-pagerduty=true \
  --set-string tags.agent-aws=true

Pick only the agents you need

Enable only the components you want (no basic/complete profile):

helm install ai-platform-engineering oci://ghcr.io/cnoe-io/charts/ai-platform-engineering \
  --version 0.2.32 \
  --namespace ai-platform-engineering \
  --create-namespace \
  --set-string tags.agent-backstage=true \
  --set-string tags.agent-slack=true \
  --set-string tags.rag-stack=true

Note: For the RAG stack use tags.rag-stack=true. For other agents use tags.agent-<name>=true (e.g. tags.agent-github=true). See Chart components for the full list.

Use a values file

If you cloned the repo, you can create a values.yaml and install from it:

# values.yaml
tags:
  basic: true
  agent-aws: true

helm install ai-platform-engineering oci://ghcr.io/cnoe-io/charts/ai-platform-engineering \
  --version 0.2.32 \
  --namespace ai-platform-engineering \
  --create-namespace \
  --values values.yaml

Chart components

Core components

Component	Version	Description
supervisor-agent	0.1.1	Multi-agent orchestration and coordination
slim	v0.1.8	AGNTCY Slim dataplane service
slim-control-plane	v0.1.3	AGNTCY Slim control plane
rag-stack	0.0.1	RAG (Retrieval-Augmented Generation) stack
backstage-plugin-agent-forge	0.1.0	Backstage plugin for agent management

Agent components

All agent subcharts use version 0.2.2. Enable with tags.agent-<name>=true (or tags.rag-stack=true for RAG).

Agent	Tag	Profiles	Description
agent-argocd	`agent-argocd`	basic, complete	ArgoCD GitOps integration
agent-aws	`agent-aws`	complete	AWS cloud resource management
agent-backstage	`agent-backstage`	basic, complete	Backstage developer portal
agent-confluence	`agent-confluence`	complete	Confluence documentation
agent-github	`agent-github`	basic, complete	GitHub repos and workflows
agent-jira	`agent-jira`	complete	Jira issue tracking
agent-komodor	`agent-komodor`	complete	Komodor Kubernetes troubleshooting
agent-pagerduty	`agent-pagerduty`	complete	PagerDuty incidents
agent-slack	`agent-slack`	complete	Slack messaging
agent-splunk	`agent-splunk`	complete	Splunk log analytics
agent-webex	`agent-webex`	complete	Webex collaboration
rag-stack	`rag-stack`	complete	RAG knowledge base and embeddings

Other installation options

Install from a local chart (after clone)

# From repo root
helm pull oci://ghcr.io/cnoe-io/charts/ai-platform-engineering --version 0.2.32

helm install ai-platform-engineering ai-platform-engineering-0.2.32.tgz \
  --namespace ai-platform-engineering \
  --create-namespace \
  --set-string tags.basic=true

ArgoCD

To deploy the chart via ArgoCD, use an Application manifest. Example:

apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: ai-platform-engineering
  namespace: argocd
  finalizers:
    - resources-finalizer.argocd.argoproj.io
spec:
  project: default
  sources:
    - chart: ai-platform-engineering
      repoURL: ghcr.io/cnoe-io/charts
      targetRevision: 0.2.32
      helm:
        parameters:
          - name: tags.basic
            value: "true"
          - name: tags.agent-aws
            value: "true"

Enable RAG stack only

helm install ai-platform-engineering oci://ghcr.io/cnoe-io/charts/ai-platform-engineering \
  --version 0.2.32 \
  --namespace ai-platform-engineering \
  --create-namespace \
  --set-string tags.rag-stack=true

Enable Backstage Agent Forge plugin

helm install ai-platform-engineering oci://ghcr.io/cnoe-io/charts/ai-platform-engineering \
  --version 0.2.32 \
  --namespace ai-platform-engineering \
  --create-namespace \
  --set-string tags.backstage-agent-forge=true

Troubleshooting

Pods not starting

Check resources: kubectl describe pod <pod-name> -n ai-platform-engineering
Verify secrets are created and correct
Confirm image pull permissions and cluster resource quotas

Agent authentication failures

Ensure required agent secrets exist for the agents you enabled
Verify credentials are valid and have the right permissions

Chart install or upgrade fails

Ensure Kubernetes version is 1.28+
Confirm namespace and RBAC allow the chart to create resources
For local development, run helm dependency update in the chart directory if you are building from source

Next steps

Configure Agent Secrets — Create secrets for GitHub, ArgoCD, LLMs, etc.
Configure LLMs for EKS — LLM provider and API keys
Run with KinD — One-command local setup with setup-caipe.sh
Run with EKS — Create an EKS cluster and deploy CAIPE

Step 1: Clone the repository (optional but recommended)​

Step 2: Prerequisites​

Step 3: Install the chart​

Basic installation (ArgoCD, Backstage, GitHub agents)​

Complete profile (all agents)​

Step 4: Verify the deployment​

Customising the deployment​

Add specific agents​

Pick only the agents you need​

Use a values file​

Chart components​

Core components​

Agent components​

Other installation options​

Install from a local chart (after clone)​

ArgoCD​

Enable RAG stack only​

Enable Backstage Agent Forge plugin​

Troubleshooting​

Pods not starting​

Agent authentication failures​

Chart install or upgrade fails​

Next steps​