How to Set Up OpenClaw: Complete Guide (2026)

TL;DR: This guide covers three ways to get OpenClaw running — from 5-minute automated deploy to full manual setup. Includes free Kimi setup on Nvidia for zero API costs.

What is OpenClaw?

OpenClaw is an open-source AI assistant gateway that runs on your own server. Unlike ChatGPT or Claude where your conversations train their models, OpenClaw keeps everything on your infrastructure.

Private: Your data never leaves your machine
Flexible: Connect multiple AI providers (Claude, GPT, Gemini, Kimi)
Multi-platform: Telegram, Discord, WhatsApp, Email, Slack
Open source: github.com/openclaw/openclaw

Method 1: 5-Minute Deploy (Recommended)

The fastest way to get OpenClaw running without touching a terminal.

What you need:

A Telegram account (for bot integration)
An AI provider API key (we'll use free Kimi)

Steps:

Go to open-claw.space
Choose your AI model (recommend: Kimi via Nvidia for free credits)
Paste your Telegram bot token
Click "Deploy"
Wait 5 minutes

Behind the scenes:

Server spins up (Ubuntu VPS)
OpenClaw installs automatically
Security configured (UFW firewall, fail2ban, non-root user)
Encrypted API keys
Telegram webhook connected

Result: Your OpenClaw instance running 24/7 on your own server.

Cost: $49/month (includes deployment + $15 API credits)

Method 2: Manual Setup (1-2 Hours)

Want to understand every component? Do it yourself.

Prerequisites

A VPS (DigitalOcean, Linode, Hetzner — $5-10/month)
SSH access
Basic Linux knowledge

Step 1: Spin Up VPS (15 min)

# Ubuntu 22.04 LTS recommended
# Create droplet, get IP, SSH in
ssh root@your-vps-ip

Step 2: Install Node.js (5 min)

curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt-get install -y nodejs
node --version  # v20.x.x

Step 3: Install OpenClaw (7 min)

git clone https://github.com/openclaw/openclaw.git
cd openclaw
npm install
npm run build

Step 4: Configure Everything (10 min)

# Copy example config
cp config.example.json config.json

# Edit with your settings
nano config.json

Key settings:

gateway.host: Set to localhost (security)
telegram.botToken: From @BotFather
ai.provider: Choose your provider
ai.apiKey: Your API key

Step 5: Set Up AI Provider (10 min)

Option A: Kimi (Free via Nvidia)

Go to build.nvidia.com/moonshot
Create free account
Generate API key
Paste into config

Free tier: 50,000 tokens/day

Option B: Anthropic Claude

console.anthropic.com
Add billing
Generate API key

Option C: OpenAI

platform.openai.com
Add billing
Generate API key

Step 6: Connect Telegram (10 min)

Message @BotFather → /newbot
Name your bot
Copy token to config
Start OpenClaw: npm start
Message your bot — it should reply

Step 7: Security Hardening (10 min)

# Firewall
sudo ufw default deny incoming
sudo ufw default allow outgoing
sudo ufw allow 22/tcp  # SSH
sudo ufw enable

# fail2ban
sudo apt install fail2ban
sudo systemctl enable fail2ban

# Non-root user
sudo useradd -m openclaw
sudo usermod -aG sudo openclaw
# Run OpenClaw as this user, not root

Step 8: Debug Why Nothing Works (?? min)

This is where most people get stuck.

Common issues:

Telegram webhook URL wrong
Firewall blocking requests
API key permissions
Node version mismatch
Config syntax errors

Debug steps:

# Check logs
npm start 2>&1 | tee openclaw.log

# Test Telegram webhook
curl -X POST https://api.telegram.org/bot<TOKEN>/getMe

# Check if gateway is listening
netstat -tlnp | grep 3000

Total time: 1-2 hours if nothing goes wrong. Days if you're learning as you go.

Method 3: Docker Deploy (30 min)

For those who know Docker.

# Pull image
docker pull openclaw/openclaw:latest

# Run with config
docker run -d \
  --name openclaw \
  -v $(pwd)/config.json:/app/config.json \
  -p 3000:3000 \
  openclaw/openclaw:latest

Pros: Isolated, reproducible
Cons: Still need to configure everything manually

Which Method Should You Choose?

Method	Time	Cost	Effort
5-Min Deploy	5 min	$49/mo	Zero
Manual	1-2 hrs (or days)	$5-10/mo VPS + API costs	High
Docker	30 min	Same as manual	Medium

My recommendation: If you value your time, use the 5-minute deploy. If you want to learn OpenClaw's internals, do it manually once — then you'll appreciate the automation.

Free AI: Kimi on Nvidia Explained

Most guides skip this. Here's the exact setup.

Why Kimi?

50,000 free tokens/day
Good quality (Moonshot's model)
No credit card required

Setup steps:

Go to build.nvidia.com/moonshot
Sign up with email
Click "Get API Key"
Copy key starting with nvapi-...
In config.json:

{
  "ai": {
    "provider": "nvidia",
    "apiKey": "nvapi-xxxxxxxxxxxxxxxxxxxxxxxxxx",
    "model": "kimi-k2.5"
  }
}

Restart OpenClaw

Monitoring usage: Dashboard shows daily token usage. If you hit 50k/day, switch to another free tier or add paid credits.

Troubleshooting Common Errors

"Webhook failed"

Check webhook URL in config
Ensure VPS IP is public
Verify SSL certificate (letsencrypt)

"API key invalid"

Check key has correct permissions
Ensure billing is active (for paid providers)
Verify key format (some need "Bearer" prefix)

"Gateway connection refused"

Check if OpenClaw is running: pm2 status or ps aux | grep openclaw
Verify port 3000 isn't blocked
Check firewall rules

"Bot not responding"

Message @BotFather, ensure bot isn't blocked
Check webhook is set: https://api.telegram.org/bot<TOKEN>/getWebhookInfo
Verify your VPS can reach Telegram servers

"Rate limited"

You're hitting API limits
Add API key for additional provider
Check token usage dashboard

Next Steps After Setup

Add integrations: Connect Discord, WhatsApp, Email
Configure skills: Enable browser automation, web search
Set up cron: Schedule automated tasks
Secure: Add 2FA to your VPS
Backup: Export your config regularly

Resources

OpenClaw Docs: docs.openclaw.ai
Community: github.com/openclaw/openclaw/discussions
Nvidia Build: build.nvidia.com

Summary

Three ways to OpenClaw:

Fast: 5-minute deploy at open-claw.space
Manual: 1-2 hours learning every component
Docker: 30 minutes if you know containers

The fastest route to private AI? 5-minute deploy. The most educational? Do it manually once, then automate.

Want to skip the headache? I built the 5-minute deploy after spending 3 days on manual setup so you don't have to.

Last updated: February 2026