There are two ways to set up OpenClaw in 2026. The hard way takes hours. The easy way takes minutes.
The Hard Way (Traditional Setup)
bash
# Install Node.js 24
# Install Docker
# Clone the repo
git clone https://github.com/openclaw/openclaw
# Configure environment variables
# Set up gateway authentication
# Configure model provider API keys
# Deal with port conflicts
# Fix Docker networking issues
# Debug "gateway not starting" errors
# ... 2 hours later, maybe it worksCommon problems people hit:
- โNode version mismatch
- โDocker port 5432 already in use
- โGateway crashes on startup
- โ"control ui requires device identity" errors
- โWebSocket connection refused
- โAPI key configuration confusion
The Easy Way (AgentCub)
- 1.Go to agentcub.live
- 2.Enter your email and create a PIN
- 3.Click "OpenClaw" to create your agent
- 4.Wait ~2 minutes for your agent to start
- 5.Click "Open Agent Control UI" โ you're in
That's it. No Docker. No CLI. No API keys. GPT-4.1 is pre-configured.
What You Get
- โFull OpenClaw Control UI (chat, files, tools, skills)
- โAzure OpenAI GPT-4.1 ready to use
- โHTTPS with Let's Encrypt
- โDedicated container (not shared)
- โWeb search via SearXNG
- โ1 CPU, 2GB RAM
Things to Try First
After your agent is running, try these prompts:
- 1."Write a comparison of Python vs Rust for web APIs with a table"
- 2."Draft 3 versions of a project delay email to a client"
- 3."Create a business plan for a pet-sitting app called PawPal"
Each one shows off what makes OpenClaw different from ChatGPT โ it doesn't just answer, it does the work.