What You'll Need

Step 1: Install OpenClaw (2 minutes)

Open your terminal and run the installer. It handles everything automatically — detects your system, installs what's needed, and launches the setup wizard.

What's a terminal? How do I open it?

The terminal is an app on your computer where you can type commands. You'll only need it for this one step — after that, you'll talk to your agent through a normal chat interface.

  • Mac: Press Cmd + Space, type Terminal, and hit Enter.
  • Windows: Click the Start menu, type PowerShell, and open it.
  • Linux: Press Ctrl + Alt + T (or search for "Terminal" in your apps).

Once it's open, you'll see a blinking cursor. That's where you paste the install command below.

On Mac or Linux, paste this and press Enter:

curl -fsSL https://openclaw.ai/install.sh | bash

On Windows, open PowerShell and paste this:

iwr -useb https://openclaw.ai/install.ps1 | iex
What does this command actually do?

It downloads the OpenClaw installer from their website and runs it. The installer:

  1. Checks what kind of computer you have
  2. Installs any missing pieces it needs (like Node.js, a programming runtime)
  3. Installs OpenClaw itself
  4. Launches the setup wizard so you can configure everything

It's the same process you'd go through installing any app — just done through the terminal instead of clicking "Download" on a website.

That's the last time you need to paste a command. The installer launches the onboarding wizard automatically.

Step 2: Walk Through the Setup Wizard (5 minutes)

After installation, the onboarding wizard runs automatically. It walks you through each choice with clear prompts — no guessing required. Here's what to expect and how to choose:

Choose where to run your Gateway

The Gateway is the engine that keeps your agent running. Think of it like a home base — it's the program on your computer that your AI agent "lives" in.

The wizard asks where to run it:

Our recommendation: Start local. You can always move to a server later.

Pick your AI provider

The wizard shows you a list of AI providers. Depending on the option you choose, you'll either sign in with OAuth or paste an API key.

What's an API key and how do I get one?

An API key is like a password that lets OpenClaw talk to an AI service on your behalf. It's how your agent gets its "brain." You create one on the provider's website, copy it, and paste it into the setup wizard.

It costs money per use when you use API billing. Some OpenClaw setups can also use provider OAuth, depending on the model and plan you choose.

Option A: OpenAI Codex OAuth — recommended for most people

  1. Choose OpenAI Codex in OpenClaw
  2. Sign in with your ChatGPT account when prompted
  3. Authorize OpenClaw to use your Codex access
  4. Pick GPT-5.4 during model selection

Codex is currently included with ChatGPT Free, Go, and all paid tiers, with higher rate limits on paid plans.

Option B: Anthropic API

  1. Go to console.anthropic.com
  2. Create a free account (email + password)
  3. Click API Keys in the left sidebar
  4. Click Create Key
  5. Give it a name (anything — "OpenClaw" works)
  6. Copy the key that appears (it starts with sk-ant-)
  7. Add a payment method under Billing (you'll only be charged for what you use)

Option C: OpenAI API key

  1. Go to platform.openai.com/api-keys
  2. Create a free account or sign in
  3. Click Create new secret key
  4. Copy the key (it starts with sk-)
  5. Add a payment method under Billing

Option D: Google (Gemini)

  1. Go to aistudio.google.com/apikey
  2. Sign in with your Google account
  3. Click Create API Key
  4. Copy the key

Keep your API key private — treat it like a password. Don't share it or post it publicly.

Don't overthink the provider choice — you can change it later or even use multiple at once.

Pick a model

After choosing a provider, you'll pick a specific model:

Starter (recommended): GPT-5.4 via OpenAI Codex OAuth. Cost-efficient. Sign in with your ChatGPT account in OpenClaw.

Pro (for deeper orchestration): Claude Opus. Best reasoning quality. More expensive per turn, so upgrade when you start hitting limits on strategic or architectural work.

GPT-5.4 runs via OpenAI Codex. Codex is currently included with ChatGPT Free, Go, and all paid tiers; paid tiers get higher rate limits. API key auth is also supported for pay-as-you-go usage.

Connect a messaging channel (optional)

This step lets your agent message you on your phone — like having a team member who can text you when something needs attention. You can skip this and add it later, but Telegram is the easiest to start with.

How to set up Telegram (step by step)

Telegram is a free messaging app (like WhatsApp or iMessage). Your agent uses a "bot" account on Telegram to send you messages. Here's how to set it up:

  1. Install Telegram on your phone if you haven't already (free on the App Store and Google Play)
  2. Open Telegram and search for @BotFather — this is Telegram's official tool for creating bots
  3. Send the message /newbot to BotFather
  4. Pick a display name for your bot (e.g., "My AI Agent")
  5. Pick a username ending in "bot" (e.g., "myagent_bot")
  6. BotFather will reply with a token — a long string of numbers and letters. Copy this.
  7. Back in the OpenClaw wizard, paste the token when it asks for your Telegram bot token
  8. Open a chat with your new bot on Telegram and send any message. If it replies, you're connected!

That's it. Your agent can now send you alerts, status updates, and summaries right to your phone.

Other supported channels include WhatsApp, Discord, iMessage, Slack, and more. You can add any of them later.

Install as a background service (recommended)

The wizard asks if you want the Gateway to start automatically. Say yes.

What does "background service" mean?

Without this, your agent only runs while your terminal window is open. With it, your agent runs quietly in the background — even after you close the terminal, and it starts again automatically when you restart your computer.

Think of it like Spotify or Dropbox — apps that run in the background without you having to open them every time. That's what this does for your agent.

Step 3: Verify Everything Works (1 minute)

After onboarding finishes, it opens a chat window where you can talk to your agent immediately. If your agent responds to a message, everything is working.

You can also open the dashboard anytime:

openclaw dashboard

This opens a web interface in your browser where you can chat, see your agent's status, and manage settings.

Step 4: Teach Your Agent Who You Are (15 minutes)

This is the step most people skip — and why most AI agents feel useless. Without context, your agent is just a chatbot. With context, it becomes a team member that knows your business.

Open the chat with your agent (through the dashboard or Telegram) and introduce yourself. Here's what to say:

Say this to your agent:

"I want you to set up your workspace so you remember who I am between sessions. Create the files you need. Here's what to know about me: My name is [your name]. I'm in [your timezone]. I run [describe your business or what you do]. I prefer [how you like to communicate — brief updates? detailed reports? casual tone?]. My current focus is [your main project or goal right now].

Also set up a personality for yourself — be genuinely helpful, have opinions, figure things out before asking me, and surface blockers immediately instead of sitting on problems."

Your agent will create the files it needs to remember you — your preferences, your business context, and how you like to work. You don't need to write or organize any of this yourself.

How does my agent remember things between conversations?

AI agents don't have permanent memory by default — each conversation starts fresh. OpenClaw solves this with a "workspace": a folder of simple text files that your agent reads at the start of every session.

When you tell your agent about yourself, it saves that information into these files. Next time it wakes up, it reads them first — so it always knows who you are, what you're working on, and how you like things done.

Think of it like leaving a note on your desk for a coworker who comes in on a different shift. The workspace is that note.

Step 5: Set Up Automatic Check-Ins (5 minutes)

This is where your agent stops being a chatbot and starts being an operator. A "heartbeat" is an automatic check-in — your agent wakes up on a schedule (say, every hour), looks at your projects and tasks, and either takes action or stays quiet if nothing needs attention.

You don't have to be in a conversation for this to happen. It runs on its own, like a team member checking in throughout the day.

Say this to your agent:

"Set up an automatic heartbeat that runs every hour. When it fires, check if there are any blockers, upcoming deadlines, or tasks that need attention. If something's urgent, message me on Telegram. If everything's fine, just note that you checked and move on."

Your agent handles the configuration. You'll start getting proactive updates without asking for them — that's the difference between a chatbot and an operator.

Step 6: Give It Something Real to Do (5 minutes)

The best way to test your agent is with a real task, not a toy example. Here are some good first assignments:

Morning briefing:

"Set up a scheduled task that runs every morning at 9 AM. Check my current projects for any blockers or upcoming deadlines, and send me a 2-3 bullet summary on Telegram. If nothing needs attention, just say 'all clear.'"

Research task:

"Research [topic relevant to your business] and write up a summary of what you find. Save it somewhere I can review it later."

Content help:

"Draft a blog post about [topic]. I want it to be [tone/length/audience]. Save it as a draft and let me know when it's ready for review."

Start with one task. When you see it deliver, you'll start thinking of ten more.

What's Next

Over the next few days, you'll notice your agent building daily logs, learning from mistakes, and getting better at anticipating what you need.

Day 2 priorities:

The key insight: the skill you're building is how to communicate with your agent effectively. The better you describe what you want, the better the results. That's a skill that transfers to every AI tool, not just OpenClaw.

Want the full guide?

The OpenClaw Playbook covers advanced workflows, multi-agent patterns, content automation, and everything that happens after day one.

Get the Playbook →