Skip to main content

​
Prerequisites

  • macOS or Linux
  • Node.js 18+ (for CLI)
  • Lima (macOS only): brew install lima

​
Step 1: Install

# Install Lima (if not already installed)
brew install lima

# Install CLI
npm install -g @orpheus/cli

# Start VM (first time takes ~3 min)
orpheus vm start

# Verify
orpheus status

​
Step 2: Create Agent

Create a folder with two files: agent.yaml 
name: hello-agent
runtime: python3
module: agent.py
entrypoint: handler

memory: 512
timeout: 60

scaling:
  min_workers: 1
  max_workers: 3
agent.py 
def handler(input_data):
    name = input_data.get('name', 'World')
    return {
        'message': f'Hello, {name}!',
        'status': 'success'
    }

​
Step 3: Deploy

orpheus deploy .
First deploy takes ~30s (builds runtime). Subsequent deploys take ~2s.

​
Step 4: Run

orpheus run hello-agent '{"name": "Orpheus"}'
Output: 
{
  "message": "Hello, Orpheus!",
  "status": "success"
}
No cold start - worker was already warm!

​
Step 5: Verify

# List agents
orpheus list

# Check stats
orpheus stats hello-agent

# View execution history
orpheus runs hello-agent

​
What Just Happened

  1. Worker stayed warm - No cold start on second request
  2. Execution logged - Every request tracked in ExecLog
  3. Isolated container - Ran in sandboxed runc container

​
Next Steps

Core Concepts

Learn how autoscaling works

Examples

See more agent patterns