Setting Up OpenClaw — A Step-by-Step Guide for Beginners

Setting Up OpenClaw: A Beginner's Guide to Your First Agent

You've decided OpenClaw is for you. Now what? This guide walks you through everything from zero to your first working agent in about 30 minutes.

No deep technical knowledge required. If you can use a terminal and copy-paste commands, you can set up OpenClaw.

Two Ways to Get Started

Option 1: myHermy (Easiest, 5 minutes)

Use myHermy, the managed hosting platform:

Pros:

One-click deployment
No server management
Automatic backups
Professional support
Updates handled for you

Cons:

Monthly cost ($20-200 depending on plan)
Data on myHermy's servers (though encrypted)

Best for: Non-technical users, businesses, first-time experimenters

Option 2: Self-Hosted (More control, 30-60 minutes)

Install OpenClaw on your own server or laptop:

Pros:

Free (after initial setup)
Complete data privacy
Full customization
No dependencies on external services

Cons:

Requires technical setup
You manage servers and updates
Requires some upkeep

Best for: Technical users, privacy-sensitive applications, long-term users

This guide covers both. Choose one.

Path 1: myHermy Setup (5 Minutes)

Step 1: Create Account

Visit myhermy.ai and click "Sign Up":

Email: your-email@example.com
Password: strong-password-here
Name: Your Name

Step 2: Choose a Plan

Plans available:

Plan	Cost	Agents	Compute
Starter	$20/month	1	Small
Professional	$60/month	5	Medium
Enterprise	Custom	Unlimited	Large

For starting: choose Starter.

Step 3: Deploy Your First Agent

Click "Create Agent". You'll see:

Agent Name: "My First Agent"
Type: "General Purpose"
Capabilities: (checkboxes for email, browser, etc.)

Select these capabilities:

✅ Browser Automation
✅ File Operations
✅ Code Execution
❌ Email (optional, set up later if needed)

Click "Deploy".

Step 4: Get Your API Key

In Settings → API Keys, copy your key. You'll use this to talk to your agent from other tools.

Step 5: Test Your Agent

Click "Chat" and try:

You: "What time is it right now?"

Agent: "The current time is [timestamp]"

Congratulations! Your agent is running.

That's it for myHermy. You're done.

Path 2: Self-Hosted Setup (30-60 Minutes)

Prerequisites

Before starting, make sure you have:

A laptop or desktop: Any modern computer (Windows, Mac, Linux)
Terminal access: Comes built-in (Terminal on Mac/Linux, PowerShell on Windows)
Git: Download from git-scm.com
Node.js: Download v18+ from nodejs.org

Total setup time for these: ~10 minutes

Step 1: Open Terminal

Mac/Linux: Open Terminal app

Windows: Open PowerShell (right-click → "Run as Administrator")

Step 2: Clone OpenClaw

Copy-paste this:

git clone https://github.com/bfzli/openclaw.git
cd openclaw

This downloads OpenClaw code to your computer.

Step 3: Install Dependencies

Copy-paste this:

npm install

This installs all the packages OpenClaw needs. Takes 2-3 minutes.

Step 4: Configure Environment

Copy-paste this:

cp .env.example .env

This creates a configuration file. Open it in your text editor (Notepad, VS Code, etc.):

# On Mac/Linux:
nano .env

# Or open in your editor:
code .env

Find these lines and update them:

# LLM Configuration - Choose one:

# Option A: Use Claude (requires Anthropic API key from https://console.anthropic.com)
LLM_PROVIDER=anthropic
ANTHROPIC_API_KEY=sk-ant-xxxxx

# Option B: Use ChatGPT (requires OpenAI API key from https://platform.openai.com/api-keys)
LLM_PROVIDER=openai
OPENAI_API_KEY=sk-proj-xxxxx

# Option C: Use Gemini (requires Google API key from https://aistudio.google.com/apikey)
LLM_PROVIDER=google
GOOGLE_API_KEY=xxxxx

# If you want to use local models (no API key needed), keep this:
# LLM_PROVIDER=ollama
# OLLAMA_HOST=localhost:11434

Getting an API Key:

The easiest option is Claude (Anthropic):

Go to https://console.anthropic.com
Sign up for a free account
Go to "API Keys"
Click "Create Key"
Copy the key
Paste into .env as ANTHROPIC_API_KEY=sk-ant-xxxxx

You'll need to add a payment method and purchase credits to use the API. Start small — even a few dollars of credits is plenty to experiment.

Save the file.

Step 5: Start OpenClaw

Copy-paste this:

npm run gateway

You should see:

Gateway starting on port 3000...
✓ Database initialized
✓ Skills loaded
✓ Memory system ready
✓ Gateway listening on http://localhost:3000

Leave this terminal window open.

Step 6: Open Another Terminal

You need a second terminal window to interact with OpenClaw.

Mac/Linux: Cmd+Tab to another Terminal window

Windows: Open another PowerShell window

Navigate to the OpenClaw folder:

cd openclaw

Step 7: Test Your Agent

Copy-paste this:

npm run task "What's the weather like right now?"

The agent will think for a moment, then respond:

Agent: "I don't have access to real-time weather data,
but I can help you search for it online if you'd like."

Congratulations! Your agent works.

Your First Real Task

Now try something more useful:

Task: Get Information About a Topic

npm run task "Research the top 5 AI agents being developed in 2026.
Focus on capabilities and who's funding them."

The agent will:

Understand the request
Browse the web for information
Compile findings
Return a summary

Takes 30-60 seconds.

Task: Create a File

npm run task "Create a file called 'shopping-list.txt' with these items:
- Milk
- Eggs
- Bread
- Coffee"

The agent creates the file. Check your project folder.

Task: Write Code

npm run task "Write a Python script that generates 10 random numbers
between 1 and 100 and calculates their average."

The agent writes code, saves it, runs it, and shows the output.

Configuring Your Agent

You probably want to customize your agent. Create a file called my-agent.config.json:

{
    "name": "My Personal Assistant",
    "type": "general",

    "capabilities": {
        "browser": true,
        "code_execution": true,
        "file_operations": true,
        "email": true,
        "calendar": true
    },

    "preferences": {
        "response_style": "professional",
        "verbosity": "concise",
        "language": "English"
    },

    "limits": {
        "max_api_calls_per_task": 10,
        "max_file_size_mb": 100,
        "timeout_seconds": 300
    }
}

Then load it:

npm run task:with-config my-agent.config.json "Your task here"

Adding Email (Optional)

To let your agent access email:

Gmail:

Go to myaccount.google.com/security
Enable "2-Step Verification"
Create an "App Password" (not your regular password)
Copy the 16-character password
In .env, add:

EMAIL_PROVIDER=gmail
GMAIL_EMAIL=your-email@gmail.com
GMAIL_APP_PASSWORD=xxxx xxxx xxxx xxxx

Restart OpenClaw:

# Stop the currently running gateway (Ctrl+C in that terminal)
# Then:
npm run gateway

Now try:

npm run task "What's in my email inbox?"

Stopping OpenClaw

When you're done, stop the Gateway:

In the terminal where it's running, press:

Ctrl+C

Next Steps

1. Read the Documentation

Visit openclaw.io/docs for detailed docs.

2. Explore Skills

OpenClaw has "skills" (modules for specific capabilities). See available skills:

npm run skills:list

Install one:

npm run skills:install email-advanced

3. Create Scheduled Tasks

Schedule tasks to run automatically:

npm run schedule "Every morning at 9 AM, send me a summary of yesterday's emails"

4. Set Up Integrations

Connect OpenClaw to:

Slack (message your agent from Slack)
Telegram (message your agent from Telegram)
Discord (use agent in Discord servers)

See integration docs.

5. Deploy to a Server

When you're ready to leave this running 24/7, deploy to a VPS:

Self-Hosting OpenClaw Guide

Troubleshooting

"API key not valid"

Problem: You're getting authentication errors
Solution:
1. Copy your API key again carefully (no extra spaces)
2. Paste it in .env
3. Save the file
4. Restart OpenClaw (Ctrl+C, then npm run gateway)

"npm: command not found"

Problem: Terminal doesn't recognize npm
Solution:
1. Restart terminal
2. Reinstall Node.js from nodejs.org
3. Restart computer if still failing

"Port 3000 already in use"

Problem: Something else is using port 3000
Solution 1: Quit other apps using that port
Solution 2: Use a different port:
  GATEWAY_PORT=3001 npm run gateway

"Agent not responding"

Problem: You run a task and it hangs
Solution:
1. Wait up to 2 minutes (first run downloads models)
2. If still nothing, stop (Ctrl+C) and restart
3. Check API key is correct
4. Check internet connection works

"File not found"

Problem: Agent can't find files it created
Solution:
1. Check file is in the right folder
2. Use absolute paths in requests
3. Check file permissions

Common First-Time Mistakes

❌ Don't: Put API key in version control

✅ Do: Keep API key in .env which is in .gitignore

❌ Don't: Run multiple gateways at once

✅ Do: Use one gateway and multiple agents connected to it

❌ Don't: Give agent access to everything

✅ Do: Start with minimal capabilities, add as needed

❌ Don't: Test with real data

✅ Do: Test with sample/dummy data first

❌ Don't: Max out API calls immediately

✅ Do: Use free tier, understand costs, scale gradually

Cost Expectations

Free Tier

Claude API: Pay-as-you-go (start with a small credit purchase) OpenAI API: Pay-as-you-go Google Gemini: Free tier available Local models via Ollama: Completely free

Cost: Near $0 to start with free tiers and local models

Low Usage (<100M tokens/month)

OpenClaw tasks: ~$10/month in API costs

Total: $10-15/month

Medium Usage (1B tokens/month)

OpenClaw tasks: ~$100/month in API costs OR $70/month self-hosted

Total: $70-100/month

Heavy Usage (10B+ tokens/month)

Self-hosted is cheaper (breaks even)

Total: $150-500/month depending on infrastructure

What to Do Now

Choose: myHermy or self-hosted?
Set up: Follow the right guide above
Test: Run a few test tasks
Integrate: Connect to your systems (Slack, email, etc.)
Automate: Schedule recurring tasks
Expand: Add more capabilities as needed

You now have a personal AI agent working for you. That's powerful.

Conclusion: You're Ready

Setting up OpenClaw isn't complicated. Whether you choose myHermy for simplicity or self-hosted for control, you can have your first agent running in under an hour.

From there, the possibilities expand. Your agent learns your patterns, integrates with your systems, and works autonomously.

The future of work isn't distant. It's on your computer right now.

Welcome to autonomous AI.

Setting Up OpenClaw - A Step-by-Step Guide for Beginners

Setting Up OpenClaw: A Beginner's Guide to Your First Agent

Two Ways to Get Started

Option 1: myHermy (Easiest, 5 minutes)

Option 2: Self-Hosted (More control, 30-60 minutes)

Path 1: myHermy Setup (5 Minutes)

Step 1: Create Account

Step 2: Choose a Plan

Step 3: Deploy Your First Agent

Step 4: Get Your API Key

Step 5: Test Your Agent

Path 2: Self-Hosted Setup (30-60 Minutes)

Prerequisites

Step 1: Open Terminal

Step 2: Clone OpenClaw

Step 3: Install Dependencies

Step 4: Configure Environment

Step 5: Start OpenClaw

Step 6: Open Another Terminal

Step 7: Test Your Agent

Your First Real Task

Task: Get Information About a Topic

Task: Create a File

Task: Write Code

Configuring Your Agent

Adding Email (Optional)

Stopping OpenClaw

Next Steps

1. Read the Documentation

2. Explore Skills

3. Create Scheduled Tasks

4. Set Up Integrations

5. Deploy to a Server

Troubleshooting

"API key not valid"

"npm: command not found"

"Port 3000 already in use"

"Agent not responding"

"File not found"

Common First-Time Mistakes

Cost Expectations

Free Tier

Low Usage (<100M tokens/month)

Medium Usage (1B tokens/month)

Heavy Usage (10B+ tokens/month)

What to Do Now

Conclusion: You're Ready