Setup & Deployment
How to Use OpenClaw: A Complete Beginner's Guide
6 min read · Updated 2026-03-01
By DoneClaw Team · We run managed OpenClaw deployments and write from hands-on production experience.
OpenClaw has exploded in popularity in early 2026, crossing 215,000 GitHub stars and making headlines everywhere from Fortune to CyberSecurity News. But if you're new to it, the question is simple: how do you actually use this thing? This guide walks you through everything — from installation to daily use — so you can get a personal AI agent running and actually working for you.
What Is OpenClaw?
OpenClaw is an open-source, self-hosted AI agent that runs on your own hardware. Unlike ChatGPT or Claude's web interface, OpenClaw gives you full control over a persistent, always-on assistant.
Think of it as a personal assistant that lives on your infrastructure and works for you around the clock.
- **Runs 24/7** on your server, Raspberry Pi, or VPS
- **Connects to your tools** — Telegram, Discord, WhatsApp, email, calendar
- **Has persistent memory** — it remembers conversations across sessions
- **Can take actions** — send emails, check your calendar, run scripts, browse the web
- **Supports any AI model** — Claude, GPT, Gemini, Ollama local models
Step 1: Install OpenClaw
You need Node.js 18+ and a machine to run it on. The most common setups: VPS ($5-10/month, most popular, always online), Raspberry Pi (cheap, runs at home), Mac mini (powerful, great for local models), or your laptop (fine for testing).
**Quick Install**
The `init` command walks you through initial setup — it creates your workspace directory and config files.
**Add an AI Model**
OpenClaw needs at least one AI model to work. The easiest option is to set an API key for Anthropic, OpenAI, or run free with Ollama locally.
For a detailed walkthrough, see our OpenClaw Setup Guide for Beginners.
npm install -g openclaw
openclaw init
# Set your API key
openclaw config set providers.anthropic.apiKey sk-your-key-here
# Or use OpenAI
openclaw config set providers.openai.apiKey sk-your-key-here
# Or run free with Ollama (local)
ollama pull llama3
openclaw config set providers.ollama.baseUrl http://localhost:11434Step 2: Connect a Chat Channel
OpenClaw is designed to be used through messaging apps — you chat with it like you'd chat with a friend.
**Telegram (Recommended for Getting Started)**
1. Message @BotFather on Telegram. 2. Create a new bot and copy the token. 3. Configure OpenClaw with the config command. 4. Message your bot on Telegram — you're now talking to your AI agent.
See our Telegram setup guide for the full walkthrough.
**Other Channels**
OpenClaw also supports Discord, WhatsApp, Slack, and more. Pick whatever you use most — the experience is the same across all of them.
openclaw config set channels.telegram.token YOUR_BOT_TOKEN
openclaw gateway startStep 3: Personalize Your Agent
This is where OpenClaw gets interesting. Your workspace has key files that define who your agent is:
**SOUL.md — Your Agent's Personality**
SOUL.md defines how your agent communicates. You can set its tone, verbosity, and style.
**USER.md — About You**
USER.md tells the agent about you — your name, timezone, and preferences.
**MEMORY.md — Long-Term Memory**
Your agent automatically builds memory over time. It remembers your preferences, past conversations, and important context. This is what makes it feel like a real assistant, not a stateless chatbot.
# SOUL.md
You're direct and helpful. No corporate speak.
Keep answers brief unless asked for detail.
Use humor when it fits. Be honest.
# USER.md
- Name: Jason
- Timezone: Asia/Hong_Kong
- Preferences: Prefers budget options, values privacySkip 60 minutes of setup — deploy in 60 seconds
DoneClaw handles Docker, servers, security, and updates. Your OpenClaw agent is ready to chat in under a minute.
Deploy NowStep 4: Daily Workflows
Here's how people actually use OpenClaw day-to-day:
**Morning Briefing**
Your agent can check your email, calendar, and news on a schedule and message you a summary. For example: "Good morning. You have 3 meetings today, 2 urgent emails from clients, and it's going to rain at 3 PM — bring an umbrella." This is configured through cron jobs and heartbeats.
**Quick Tasks via Chat**
Just message your agent naturally: "What's on my calendar tomorrow?" or "Draft a reply to that email from Sarah" or "Remind me to call the dentist at 2 PM."
**Research and Writing**
OpenClaw can search the web, read documents, and produce summaries. Ask it to summarize articles, compare pricing, or write blog posts.
**Automation**
Set up recurring tasks that run automatically — check email every 30 minutes and alert you about urgent ones, monitor websites for changes, generate daily reports, or post to social media on a schedule.
Step 5: Add Skills
Skills extend what your agent can do. Think of them as plugins.
Popular skills include: Weather (forecasts and alerts), Email via Himalaya (read, search, and draft emails), Calendar via gog (Google Calendar integration), Web search (Brave Search integration), and Browser (full web browsing capability).
For more on skills, see How to Write Your First Custom OpenClaw Skill and 10 Best OpenClaw Skills on ClawHub.
# Browse available skills
openclaw skills search weather
# Install a skill
openclaw skills install weatherStep 6: Keep It Running
For your agent to work 24/7, you need it running as a background service.
OpenClaw automatically reconnects if your internet drops and handles updates gracefully.
# Using systemd (Linux/VPS)
openclaw service install
openclaw service start
# Or simply
openclaw gateway start --daemonTips for Getting the Most Out of OpenClaw
**Start simple.** Don't try to automate everything on day one. Start with chat, add one integration (email or calendar), then expand.
**Talk to it naturally.** You don't need special commands. Just chat like you would with a human assistant.
**Check your memory files.** Look at MEMORY.md periodically — it's fascinating to see what your agent remembers about you.
**Use heartbeats.** Configure periodic check-ins so your agent proactively updates you instead of waiting to be asked.
**Keep your SOUL.md updated.** As you figure out what tone and style you prefer, update the personality file. Your agent adapts immediately.
**Monitor costs.** AI API calls cost money. See our guide on how much OpenClaw actually costs and 5 ways to cut your API bill.
Common Questions
**Is OpenClaw free?** The software is free and open-source. You pay for AI model API costs and hosting (as low as $5/month on a VPS). See our cost breakdown.
**Can I use it without coding?** Basic setup requires running a few terminal commands, but daily use is entirely through chat. No coding needed once it's running. The new SaaS version requires zero technical setup.
**Is my data private?** With self-hosting, your data never leaves your machine. See our privacy assessment.
**What model should I use?** Claude and GPT-4 are the most capable. For free options, see our best free models guide.
Conclusion
Once you're comfortable with the basics, explore automation examples for inspiration, set up email integration for inbox management, try local models with Ollama for zero-cost AI, and harden your setup with our security checklist. OpenClaw gets more useful the longer you use it — its memory builds, it learns your preferences, and it becomes genuinely indispensable. The best time to start is now. Ready to deploy? Follow our complete deployment guide to get your agent running in minutes.
Skip the setup? DoneClaw deploys OpenClaw for you — $29/mo with 7-day free trial, zero configuration.
Skip 60 minutes of setup — deploy in 60 seconds
DoneClaw handles Docker, servers, security, and updates. Your OpenClaw agent is ready to chat in under a minute.
Deploy NowFrequently asked questions
How do I install OpenClaw?
You need Node.js 18+ and a machine to run it on (VPS, Raspberry Pi, Mac mini, or laptop). Run 'npm install -g openclaw' followed by 'openclaw init' to set up your workspace. Then add at least one AI model provider API key and you're ready to go.
What chat channels does OpenClaw support?
OpenClaw supports Telegram, Discord, WhatsApp, Slack, and more. Telegram is recommended for getting started — you create a bot via BotFather, add the token to your config, and start chatting with your agent directly in Telegram.
Does OpenClaw require coding skills to use?
Basic setup requires running a few terminal commands, but daily use is entirely through chat — no coding needed once it's running. You message your agent naturally through Telegram, Discord, or other channels. The SaaS version requires zero technical setup at all.
How much does it cost to run OpenClaw?
The software is free and open-source. You pay for AI model API costs (varies by usage and model choice) and hosting. A VPS costs as low as $5-10/month, or you can run it on a Raspberry Pi for zero ongoing hosting cost. Using local models via Ollama eliminates API costs entirely.
How does OpenClaw remember previous conversations?
OpenClaw has persistent memory stored in a MEMORY.md file that it automatically builds over time. It remembers your preferences, past conversations, and important context across sessions. This is what makes it feel like a real assistant rather than a stateless chatbot that forgets everything between conversations.