How to Deploy an AI Agent: Cloud, VPS, and Serverless Options

Why Deployment Matters for AI Agents

Your laptop can run an agent during testing, but for a trading bot or monitoring agent that needs to operate 24/7, you need a reliable hosting solution.

Key requirements for AI agent deployments:

• Uptime — 99.9%+ for anything handling money
• Low latency — Critical for trading bots
• Persistent storage — For agent memory and logs
• Cost efficiency — AI API calls add up

Option 1: VPS (Virtual Private Server)

Best for: Trading bots, high-frequency agents, anything requiring speed

Recommended providers:

• Hetzner (Germany/US) — Best price/performance. 2 vCPU, 4GB RAM from €4.15/month
• DigitalOcean — Easy to use, great docs. 1 vCPU, 1GB RAM from $6/month
• Vultr — High performance. Metal servers available
• AWS EC2 — Enterprise reliability, higher cost

Setup Guide (Ubuntu 22.04)

# 1. Update system
sudo apt update && sudo apt upgrade -y

# 2. Install Node.js
curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt install -y nodejs

# 3. Install PM2 for process management
npm install -g pm2

# 4. Clone your agent
git clone https://github.com/yourusername/your-agent
cd your-agent
npm install

# 5. Set environment variables
echo "OPENAI_API_KEY=sk-..." > .env

# 6. Start with PM2
pm2 start dist/index.js --name "my-agent"
pm2 save
pm2 startup  # Auto-restart on server reboot

Monitoring with PM2

pm2 status          # Check running processes
pm2 logs my-agent   # View real-time logs
pm2 monit           # CPU/memory dashboard

Option 2: Serverless (AWS Lambda, Vercel, Cloudflare Workers)

Best for: Event-driven agents, lightweight tasks, low traffic

Serverless is great for agents triggered by webhooks, scheduled tasks (cron), or API calls.

AWS Lambda Example

// handler.ts
import { runAgent } from './agent'

export const handler = async (event: AWSLambda.ScheduledEvent) => {
  console.log('Running scheduled agent task...')
  const result = await runAgent('Check crypto prices and send alert if BTC drops 5%')
  return { statusCode: 200, body: JSON.stringify({ result }) }
}

Deploy with AWS CDK:

import { Function, Runtime } from 'aws-cdk-lib/aws-lambda'
import { Schedule, Rule } from 'aws-cdk-lib/aws-events'

const agent = new Function(this, 'AgentFunction', {
  runtime: Runtime.NODEJS_20_X,
  handler: 'handler.handler',
  timeout: Duration.minutes(5),
  environment: { OPENAI_API_KEY: process.env.OPENAI_API_KEY! }
})

// Run every 5 minutes
new Rule(this, 'AgentSchedule', {
  schedule: Schedule.rate(Duration.minutes(5)),
  targets: [new LambdaFunction(agent)]
})

Cloudflare Workers (Edge)

Ideal for lightweight agents that need global low latency:

export default {
  async scheduled(event: ScheduledEvent, env: Env): Promise<void> {
    // Runs on cron schedule
    const result = await runLightweightAgent(env.OPENAI_API_KEY)
    await env.KV.put('last_result', JSON.stringify(result))
  }
}

Option 3: Docker + Self-Hosted

FROM node:20-alpine
WORKDIR /app
COPY package*.json ./
RUN npm ci --production
COPY dist ./dist
CMD ["node", "dist/index.js"]

docker build -t my-agent .
docker run -d \
  --name my-agent \
  --restart unless-stopped \
  -e OPENAI_API_KEY=sk-... \
  my-agent

Option 4: Railway / Render / Fly.io

Zero-infrastructure platforms — just push code and they handle the rest.

Railway:

npm install -g railway
railway login
railway init
railway up

Render: Connect GitHub repo, set environment variables, deploy. From $7/month.

Choosing the Right Option

| Use Case | Recommended | Cost | |----------|-------------|------| | Trading/arbitrage bot | Hetzner VPS (near exchange) | $5–20/month | | Scheduled monitoring agent | AWS Lambda or Vercel Cron | $0–5/month | | Web-facing AI app | Vercel + serverless | $0–20/month | | High-throughput agent | DigitalOcean Droplet | $10–40/month |

Security Best Practices

1. Never store API keys in code — Use environment variables
2. Use secrets managers — AWS Secrets Manager, Doppler
3. Set spending limits on OpenAI and exchange accounts
4. Monitor for anomalies — Set alerts for unusual spend or behavior
5. Use read-only API keys where possible
6. Firewall your VPS — Only open ports you need

Get Pre-Configured Deployment Templates

Our GitHub repos include Docker configs, PM2 configs, and deployment scripts for all major platforms. Check the Tools page to get started.