← Back to blog

OpenClaw Quick Start: Get Your AI Agent Running in Under 5 Minutes

| Weihao Ming

Last updated: March 2026

This is the shortest path from “I just heard about OpenClaw” to “I have a working AI agent.” Two options: cloud (3 minutes, no technical skills) or self-hosted Docker (15-30 minutes, requires terminal comfort). Pick your path.

By Weihao Ming, former AWS EC2 engineer and founder of OpenClaw Cloud.

The official OpenClaw docs are thorough. Maybe too thorough if you just want the thing running. This guide cuts everything down to the minimum steps. No theory, no architecture explanations, just commands.

If something goes wrong, the troubleshooting guide has fixes for every common error.

Path A: OpenClaw Cloud (3 minutes, no server needed)

This is the fastest way to get OpenClaw running. No terminal. No Docker. No server.

Step 1: Get an AI API key

You need an API key from at least one AI provider. Anthropic (Claude) is the most popular choice.

  1. Go to console.anthropic.com
  2. Create an account and add a payment method
  3. Go to API Keys → Create Key
  4. Copy the key (starts with sk-ant-)

Cost: ~$5-20/month for typical usage. You only pay for what you use.

Step 2: Open OpenClaw Cloud in Telegram

  1. Go to OpenClaw Cloud and join the waitlist
  2. Once approved, you will receive a Telegram bot link
  3. Open the bot in Telegram
  4. Paste your API key when prompted

Step 3: Start using it

Send a message. That is it. Your agent is live.

Try something like:

  • “Search Google for the best noise-canceling headphones under $300 and summarize the top 3 results”
  • “Go to Hacker News and find today’s top story about AI”
  • “Download the CSV from [URL] and tell me the average value in column B”

The agent has a real browser, a file system, and a terminal. It does not just generate text — it takes action.

Total time: ~3 minutes. Cost: $25/month (beta) + your API usage.

Path B: Self-hosted with Docker (15-30 minutes)

For developers who want full control. You need a machine that stays on — a VPS or a Mac Mini.

Step 1: Get an AI API key

Same as above. Anthropic, OpenAI, Google, or any supported provider.

Step 2: Get a Telegram bot token

  1. Open Telegram and message @BotFather
  2. Send /newbot
  3. Follow the prompts to name your bot
  4. Copy the bot token

Step 3: Install Docker

On Ubuntu/Debian VPS:

curl -fsSL https://get.docker.com | sh
sudo usermod -aG docker $USER
# Log out and back in for group change to take effect

On Mac: Download Docker Desktop.

Step 4: Run OpenClaw

docker pull openclaw/openclaw:latest

docker run -d --restart unless-stopped \
  --name openclaw \
  -e ANTHROPIC_API_KEY=your_key_here \
  -e TELEGRAM_BOT_TOKEN=your_bot_token_here \
  openclaw/openclaw:latest

Step 5: Set up HTTPS (required for Telegram)

Telegram webhooks require HTTPS. The fastest way:

Option A — Caddy (recommended, auto-SSL):

sudo apt install caddy

Add to /etc/caddy/Caddyfile:

yourdomain.com {
    reverse_proxy localhost:3000
}
sudo systemctl restart caddy

Option B — ngrok (for testing, no domain needed):

ngrok http 3000

Step 6: Message your bot

Open Telegram, find your bot, send a message. If everything is configured correctly, the agent will respond.

Total time: 15-30 minutes (assuming you already have a VPS). Cost: $5-24/month VPS + your API usage.

Quick reference: what to do when something breaks

ProblemFix
Cannot find module 'openclaw'Upgrade Node.js to 18+
EADDRINUSE :::3000Kill the old process: lsof -i :3000 then kill -9 <PID>
Invalid API key / 401Check key format and credits in provider dashboard
Docker container exits immediatelyRun docker logs openclaw and check for missing env vars
Telegram bot not respondingCheck webhook: curl https://api.telegram.org/bot<TOKEN>/getWebhookInfo
Out of memoryUpgrade to 4GB+ RAM or add swap

For detailed solutions to each of these, see the full troubleshooting guide.

What to do next

Once your agent is running:

  1. Learn what it can do — What Can OpenClaw Actually Do?
  2. Understand the browser — How OpenClaw’s Browser Works
  3. Set up custom skills — OpenClaw Skills Explained
  4. Configure memory — OpenClaw Memory and Context
  5. Use it for specific workflows — Research, Freelancing, Email Automation

Or just start giving it tasks. Most people learn what the agent is capable of by pushing its limits. Send it something hard and see what happens.

Cloud vs self-hosted: which is right for you?

If you are reading a quick start guide, you probably want the quickest start. That is OpenClaw Cloud — 3 minutes, no infrastructure.

If you are a developer who wants to own the stack, self-hosting gives you full control. Budget an hour for the initial setup and accept that you will be doing maintenance going forward.

For a deeper comparison, see OpenClaw Cloud vs Self-Hosting.

← Back to blog