Setup

Install & First Run

8 min

Install OpenClaw, run onboarding, verify the Gateway, and send your first message through the Control UI or a chat channel.

What You Need

  • Node.js 24 recommended, or Node.js 22.19+ for compatibility
  • A model provider credential, such as Anthropic, OpenAI, Google, OpenRouter, or another supported provider
  • Optional: a chat account such as Telegram, WhatsApp, Slack, Discord, Signal, or iMessage
  • Optional: Docker if you want sandboxed tool execution

Check your local runtime:

bash
node --version
npm --version

Install OpenClaw

The current quickest path is the installer script:

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

Windows users can use PowerShell:

powershell
iwr -useb https://openclaw.ai/install.ps1 | iex

The npm install path is still useful when you want package-manager control:

bash
npm install -g openclaw@latest
# or
pnpm add -g openclaw@latest

Windows desktop users can also start from Windows Hub, which provides setup, tray status, chat, node mode, and local MCP mode.

Run Onboarding

bash
openclaw onboard --install-daemon

Onboarding walks through:

  • model provider and credential setup
  • Gateway token/auth setup
  • workspace bootstrap
  • optional channel setup
  • service installation through launchd or systemd where supported

After onboarding, verify that the Gateway is alive:

bash
openclaw gateway status

Open the Control UI

bash
openclaw dashboard

The Control UI is the browser dashboard served by the Gateway. It can chat, inspect sessions, show channel status, stream tool activity, expose nodes, and adjust runtime settings.

Local loopback access is usually auto-approved. Remote browser access may require device pairing:

bash
openclaw devices list
openclaw devices approve <requestId>

Send Your First Message

Use the Control UI chat first. It proves the Gateway, model credential, session store, and basic runtime are working before you add channel complexity.

You can also send a one-off agent turn from the CLI:

bash
openclaw agent --agent main --message "Create a short personal assistant smoke test checklist"

Or send a channel message once a channel is configured:

bash
openclaw message send --target +15555550123 --message "Hello from OpenClaw"

Connect a Channel

Telegram is usually the fastest channel because it only needs a bot token. WhatsApp requires QR pairing and stores more local state. Slack, Discord, Signal, iMessage, Matrix, Microsoft Teams, Google Chat, Zalo, SMS, and other surfaces have their own setup constraints.

Start with status:

bash
openclaw channels status

Then add or log in to a channel:

bash
openclaw channels add
openclaw channels login --channel telegram

First Safety Pass

Before you let other people message your agent, check the basics:

bash
openclaw doctor
openclaw security audit
openclaw status --usage

Keep DM pairing enabled for public chat platforms. Treat every inbound chat message as untrusted input, especially in group rooms.

Upgrade Later

OpenClaw ships frequently. Use the updater or your package manager depending on how you installed:

bash
openclaw update
openclaw doctor

Setup as architecture

A successful first run proves four boundaries: the Gateway can accept authenticated clients, the model provider can run, the session store can persist state, and at least one surface can send messages. Everything else builds on those facts.

Continue with practice

You have finished the core ideas of Install & First Run.

If you want to turn the idea into something reusable, continue practicing on AgentWay.