Getting Started

🟢 PEMOS Beta is LIVE — Managed serverless instances available at pemos.ca →. Or self-host with the guides below.

Choose Your Deployment

🐳 Docker Compose

Fastest · $0/mo · 5 minutes

Run on your laptop or server. Encrypted disk optional. Perfect for personal use and development.

Docker Guide ↓

☁️ Serverless

Cheapest cloud · ~$22/mo · 5-10 min

Azure Container Apps, AWS Fargate, or GCP Cloud Run. Scale to zero. No cluster management.

Serverless Guide ↓

☸️ Kubernetes

Full control · $121+/mo · 2-4 hrs

AKS, EKS, GKE, or k3s. NetworkPolicy, RBAC, GitOps. For compliance-heavy enterprise.

K8s Guide ↓

Docker Compose (Recommended Start)

The fastest way to run PEMOS. Works on Linux, macOS, and Windows (WSL2).

Prerequisites

• Docker Engine 24+ (or Docker Desktop)
• 2 GB RAM available
• An LLM API key (OpenAI, Anthropic, Azure OpenAI, or local Ollama)

Step 1 — Quick Install

# Download and run the installer
curl -fsSL https://get.pemos.org | bash

# Initialize (creates config + optional encrypted disk)
pemos init

# Start the fleet
pemos start

Or manually with Docker Compose:

git clone https://github.com/eose-sre/pemos-oss.git
cd pemos-oss

# Copy and edit config
cp .env.example .env
# Edit .env: set your LLM API key, agent name, etc.

docker compose up -d

Step 2 — Open the Portal

Navigate to http://localhost:8080 in your browser. You’ll see the Gangway portal with:

• Dashboard — Fleet health and status
• Chat — Talk to your agent
• Wiki — Knowledge base

Step 3 — Configure Your Agent

Edit .env or config/openclaw.json:

# .env
LLM_PROVIDER=anthropic
LLM_API_KEY=sk-ant-...
AGENT_NAME=sencho
AGENT_MODEL=claude-sonnet-4-20250514

Restart: pemos restart or docker compose restart

Optional: Encrypted Disk

For data protection, PEMOS can use a dedicated encrypted disk image:

# Linux / WSL2 (LUKS2)
pemos init --encrypted --size 10G

# macOS (APFS encrypted)
pemos init --encrypted --size 10G --format apfs

All agent data, secrets, and chat history live on the encrypted volume. Destroy the key = destroy the data permanently.

Serverless Deployment

The cheapest cloud option. Agents scale to zero when idle — you pay nothing for sleeping agents.

Azure Container Apps (Recommended)

Proven · Cheapest · Scale-to-zero native

# Prerequisites: Azure CLI, logged in
az login

# Create resource group
az group create -n rg-pemos-dev -l canadacentral

# Deploy with Bicep (one command)
az deployment group create \
  --resource-group rg-pemos-dev \
  --template-file deploy/aca/main.bicep \
  --parameters \
    prefix=my-pemos \
    location=canadacentral \
    agentImage=eosefleetacrdev.azurecr.io/openclaw-agent:v1.0.2 \
    portalImage=eosefleetacrdev.azurecr.io/pemos-portal:v1.8.5

Cost: ~$5/mo for 1 agent, ~$22/mo for 10 agents (bursty usage).

What deploys:

• ACA Environment (Consumption plan)
• Portal container (min:1)
• Dashboard API container (min:1)
• Agent container(s) (min:0, scale-to-zero)
• Azure Key Vault (secrets)
• Azure File Share (persistent state)

AWS ECS Fargate

# Prerequisites: AWS CLI, configured
aws configure

# Deploy with CloudFormation
aws cloudformation deploy \
  --template-file deploy/aws/fargate.yaml \
  --stack-name pemos-dev \
  --parameter-overrides \
    AgentImage=your-ecr.amazonaws.com/openclaw-agent:v1.0.2 \
    PortalImage=your-ecr.amazonaws.com/pemos-portal:v1.8.5

Cost: ~$8/mo for 1 agent, ~$25-65/mo for 10 agents (bursty).

GCP Cloud Run

# Prerequisites: gcloud CLI, configured
gcloud auth login

# Deploy with gcloud
gcloud run deploy pemos-portal \
  --image=your-gar.pkg.dev/project/pemos-portal:v1.8.5 \
  --region=us-central1 \
  --min-instances=1

gcloud run deploy pemos-agent \
  --image=your-gar.pkg.dev/project/openclaw-agent:v1.0.2 \
  --region=us-central1 \
  --min-instances=0  # Scale to zero!

Cost: ~$5/mo for 1 agent, ~$20-55/mo for 10 agents (bursty).

Serverless Cost Comparison

Estimates for bursty workloads (agents active <30 min/day). Prices vary by region.

Kubernetes Deployment

Full control for compliance-heavy environments. Works on any Kubernetes cluster.

Prerequisites

• Kubernetes cluster 1.27+ (AKS, EKS, GKE, k3s, or kind)
• Helm 3.12+
• kubectl with cluster access

Step 1 — Add the Helm Repository

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

Step 2 — Install PEMOS

helm install pemos pemos/pemos-oss \
  --namespace pemos --create-namespace \
  --set gangway.host=gangway.example.com \
  --set agents.sencho.enabled=true \
  --set agents.sencho.model=anthropic/claude-sonnet-4-20250514 \
  --set agents.sencho.apiKey=sk-ant-...

This deploys: Gangway portal, OpenClaw agent(s), PostgreSQL, and Redis.

Step 3 — Verify

kubectl get pods -n pemos
# All pods should be Running within 2-3 minutes

Step 4 — Access the Portal

# Port forward for quick access
kubectl port-forward svc/gangway 8080:8080 -n pemos

# Or check Helm notes for ingress URL
helm get notes pemos -n pemos

Kubernetes Cost Comparison

Configure Your Agent

After deployment (any method), configure your agent:

LLM Providers

Messaging Channels

Connect agents to Telegram, Discord, Signal, or use the built-in web chat.

# values.yaml (Helm) or config/openclaw.json (Docker)
channels:
  telegram:
    token: "your-bot-token"
  discord:
    token: "your-bot-token"
    guildId: "your-guild-id"

When to Use What

Need help? Docs · Community · Architecture

Want managed? pemos.ca handles everything — start free, serverless, scale-to-zero.