Getting Started

Prerequisites

Kubernetes Cluster

A running K8s cluster (1.27+). AKS, EKS, GKE, k3s, or kind all work. Minimum 3 nodes recommended for production.

Helm 3

Helm 3.12+ installed on your local machine. Used to deploy the Kantai chart and manage upgrades.

kubectl

kubectl configured with cluster access. You’ll need cluster-admin or equivalent RBAC for initial setup.

Optional: A domain name with TLS cert (or cert-manager) for the Gangway portal. Works on localhost for dev.

Step 1 — Add the Helm Repository

helm repo add kantai https://charts.pemos.org
helm repo update

Step 2 — Create a Namespace

kubectl create namespace kantai

Step 3 — Install Kantai OSS

Minimal install with defaults:

helm install kantai kantai/kantai-oss \
  --namespace kantai \
  --set gangway.host=gangway.example.com \
  --set agents.sencho.enabled=true \
  --set agents.takumi.enabled=true

This deploys:

• Gangway — the web portal (Dashboard, Chat, Nava, Learn, Wiki, Governance, Tetraban)
• Sencho — your personal AI captain
• Takumi — your worker/execution agent
• OpenClaw runtime — the agent execution engine
• PostgreSQL — conversation and memory storage

Step 4 — Verify the Deployment

kubectl get pods -n kantai

All pods should reach Running within 2-3 minutes. Check logs if anything stays in Pending:

kubectl logs -n kantai deploy/gangway
kubectl logs -n kantai deploy/sencho

Step 5 — First Login to Gangway

Open your browser and navigate to your Gangway host (e.g., https://gangway.example.com).

1. The initial admin credentials are output during install — check the Helm notes:

   helm get notes kantai -n kantai
   

2. Log in and change your password immediately
3. You’ll land on the Dashboard — showing agent health, fleet status, and quick stats

Step 6 — Configure Your First Agent

From the Gangway portal:

1. Navigate to Governance → Agent Config
2. Select Sencho and configure:
   • Model: Choose your LLM backend (OpenAI, Anthropic, Azure OpenAI, local Ollama)
   • API Key: Enter your provider key (stored encrypted via K8s secrets)
   • Personality: Adjust tone, verbosity, and behavior
3. Click Save & Restart Agent

Or configure via Helm values:

agents:
  sencho:
    model: anthropic/claude-sonnet-4-20250514
    apiKey: sk-...
    tools:
      - calendar
      - notes
      - web-search

helm upgrade kantai kantai/kantai-oss -n kantai -f values.yaml

Step 7 — Connect Messaging Channels

Kantai agents can communicate over multiple channels simultaneously.

Telegram

1. Create a bot via @BotFather
2. Add the bot token to your Helm values under channels.telegram.token
3. Start chatting with your bot — Sencho responds

Discord

1. Create a Discord application and bot at discord.dev
2. Add the bot token and guild ID to channels.discord
3. Invite the bot to your server — agents join configured channels

Signal

1. Register a Signal number using signal-cli or the linked device flow
2. Configure channels.signal with your credentials
3. Agents respond to DMs and group messages

You can also use the Gangway Chat tab for a built-in web interface — no external setup needed.

Troubleshooting

Pods stuck in Pending

Check node resources: kubectl describe pod <name> -n kantai. Common cause: insufficient CPU/memory. Minimum 2 CPU / 4GB per agent pod.

Agent not responding

Verify the LLM API key is correct: kubectl logs deploy/sencho -n kantai. Look for authentication errors. Check network policies allow egress to your model provider.

Gangway 502 errors

The portal pod may still be starting. Wait 60s and retry. Check kubectl logs deploy/gangway -n kantai for port binding issues.

Database connection failures

Ensure the PostgreSQL pod is running: kubectl get pods -n kantai -l app=postgresql. Check PVC binding if using persistent storage.

Need more help? Check the Docs for full configuration reference, or visit Community for support.

Want managed hosting? pemos.ca handles infrastructure, upgrades, and support so you can focus on your agents.