Your first personal AI agent — up and running, step by step.
No coding experience required.
OpenClaw is your personal AI agent platform. Think of it as giving yourself a brilliant, always-available AI assistant that:
This is not a chatbot app. This is your AI, running on your infrastructure. Nobody else has access to your conversations or your data.
Before you start, make sure you have:
Not sure if you have Node.js? Open your Terminal and type node --version. If you see something like v24.0.0, you're good. If not, download it free at nodejs.org.
Open your Terminal (macOS/Linux) or PowerShell (Windows) and run one of these commands:
macOS / Linux
curl -fsSL https://openclaw.ai/install.sh | bash
Windows (PowerShell)
iwr -useb https://openclaw.ai/install.ps1 | iex
When it's done, verify the install worked:
openclaw --version
You should see a version number. If you do — you're in. ✅
openclaw/2.x.x or similar. Any version number means your install succeeded. If you get "command not found," close and reopen your terminal and try again.
OpenClaw has a guided wizard that configures everything. Run:
openclaw onboard --install-daemon
The wizard walks you through these sections:
Choose your AI provider and paste your API key. Recommended for beginners: Anthropic (Claude).
This is the folder where your agent stores its memory and personality. The default is ~/.openclaw/workspace — just press Enter to accept.
The Gateway is the always-running brain of OpenClaw. Default settings work perfectly — press Enter through these steps.
Connect messaging apps like WhatsApp or Telegram. You can skip this now and add channels later with openclaw configure.
When asked about installing the daemon, say Yes. This makes OpenClaw start automatically when your computer boots.
http://127.0.0.1:18789 — that's your Control UI address for the next step.
Once setup is done, launch your dashboard:
openclaw dashboard
This opens your browser to http://127.0.0.1:18789/ — the OpenClaw Control UI.
http://127.0.0.1:18789 and paste it into your browser manually.
You'll see a clean chat interface. Type something and hit send. Your AI agent will respond.
openclaw status in your terminal.
Your AI agent is live! Everything from here is about making it yours.
On your first conversation, the agent runs a short "bootstrap" ritual — it introduces itself and figures out who you are. It will ask things like:
Answer honestly — it's building its memory of you. After this, the agent creates:
IDENTITY.md — its name and personalityUSER.md — your name and preferencesBOOTSTRAP.md — the one-time setup ritual it no longer needsYour agent's home is at ~/.openclaw/workspace. These are plain text files you can read and edit anytime:
.openclaw → workspace, you'll see a collection of .md (Markdown) files. These are plain text — you can open and edit them in any text editor like TextEdit, VS Code, or even Notepad. Changes take effect on your agent's next message.
| File | What It Does |
|---|---|
AGENTS.md | How your agent operates and uses memory |
SOUL.md | Personality, tone, and values |
USER.md | Information about you |
IDENTITY.md | Your agent's name and vibe |
TOOLS.md | Notes about your specific setup |
HEARTBEAT.md | Tasks the agent checks periodically |
MEMORY.md | Long-term curated memory |
memory/ | Daily memory log files |
Pro Tip: Open SOUL.md and customize it. This is literally your agent's soul — its values, vibe, and how it talks to you. Edit it like you'd write a job description for your ideal assistant.
The Gateway is the always-running background service that powers everything. Check it with:
openclaw gateway status
openclaw gateway start to bring it back up.
Other useful commands:
openclaw gateway start # Start the gateway
openclaw gateway stop # Stop the gateway
openclaw gateway restart # Restart it
Want to chat with your AI from your phone? Connect a messaging channel — this is where OpenClaw gets really powerful.
@BotFather/newbot and follow the promptsopenclaw configure and paste the token in the Telegram section1234567890:ABCDefGhIJKlmNoPQRsTUVwxyZ. Copy the entire token including the numbers before the colon. Paste it exactly when OpenClaw prompts you — no extra spaces.
WhatsApp uses a QR code scan. Run openclaw configure, navigate to WhatsApp, and scan the QR code with your phone's camera.
openclaw configureIf you prefer a graphical setup experience, OpenClaw has a native macOS app. Here's how the first-run experience works:
When you first open the app, macOS will show a security warning (this is normal for apps outside the App Store). Click Open Anyway in System Settings → Privacy & Security.
macOS will ask if OpenClaw can find devices on your local network. Click Allow.
OpenClaw shows a security notice explaining the trust model. Read it, then proceed.
Security note: By default, OpenClaw is a personal agent — only you can access it. Don't share your gateway credentials with anyone.
Decide where your Gateway runs:
The app will request various macOS permissions. Grant what you're comfortable with:
The app can install the openclaw command-line tool globally. Say Yes — it unlocks terminal workflows and makes everything in this guide work.
After setup, the app opens a dedicated onboarding chat where your agent introduces itself and guides you through first steps. This is separate from your main chat, so it stays clean.
macOS vs CLI: Both paths lead to the same result. The macOS app is friendlier for non-technical users; the CLI gives more control. You can mix them — use the app for setup, terminal for day-to-day commands.
Here's a taste of real things you can say to your agent:
You are in full control. Paste rules like these into your first conversation:
You are my personal AI assistant. Your name is [NAME].
## Privacy Rules
- Never share my personal information externally without my approval.
- Never store sensitive data (passwords, API keys) externally.
- Always ask before accessing external services.
## Purchase Rules
- NEVER make any purchase without me saying "YES, confirm purchase."
- Always state the cost before any paid action.
## Autonomy Rules
- Always summarize what you're about to do before doing it.
- Always ask before any irreversible action.
- When uncertain, do less and ask.
Your agent will follow these rules for all future conversations.
openclaw doctor
Runs a full diagnostic and tells you exactly what's wrong.
Memory lives in files. Check:
cat ~/.openclaw/workspace/MEMORY.md
ls ~/.openclaw/workspace/memory/
openclaw onboard --reset
This resets config and credentials. Your workspace files (memory, personality) are preserved unless you add --reset-scope full.
openclaw status
Shows a full system status summary. Share this output when asking for help in the OpenClaw community.
SOUL.md to personalize the personalityUSER.md so it knows you betterHEARTBEAT.md for regular background checksSkills are plug-in abilities you can install:
openclaw skills list
Find more at clawhub.com.
Get help, share ideas, and see what others are building: discord.com/invite/clawd
| Command | What It Does |
|---|---|
openclaw dashboard | Open the Control UI in your browser |
openclaw gateway status | Check if the gateway is running |
openclaw gateway restart | Restart the gateway |
openclaw configure | Reconfigure settings / add channels |
openclaw onboard | Re-run the setup wizard |
openclaw doctor | Diagnose and fix problems |
openclaw status | Full system status overview |
openclaw skills list | See available agent skills |