How to Install OpenClaw in 2026

What is OpenClaw?

OpenClaw is an open-source AI assistant framework that connects large language models — Claude, GPT-4, Gemini — to messaging platforms like Telegram and WhatsApp. Once installed, you get a personal AI assistant that lives in your chat app, responds to messages, and can be extended with custom tools and workflows.

Installing OpenClaw means getting its gateway service running on a server and connecting it to your messaging accounts. There are three main ways to do this — we cover all of them below. If you want the short version, skip straight to Method 1.

Setup time: under 1 minute · No Docker or server required

ClawMates is a managed hosting service built specifically for OpenClaw. Instead of installing and configuring everything yourself, you enter your bot credentials and click deploy. The entire OpenClaw installation happens automatically on dedicated infrastructure.

Method 2 — Docker Self-Host

Setup time: 30–60 minutes · Requires: Linux server, Docker, AI API keys

Docker is the most reliable way to self-host OpenClaw. You get full control over the installation, configuration, and data. This method is suitable for developers comfortable with the command line and Linux server administration.

Prerequisites

Step 1 — Pull the OpenClaw Docker image

The official OpenClaw image is published on Docker Hub under the alpine/openclaw namespace:

docker pull alpine/openclaw:latest

This pulls the latest stable release. The image is based on Alpine Linux and weighs around 200MB compressed.

Step 2 — Write the OpenClaw config file

Create a configuration file at /opt/openclaw/config.json. This is the minimal config for a Telegram bot with Claude as the AI backend:

mkdir -p /opt/openclaw

cat > /opt/openclaw/config.json << 'EOF'
{
  "gateway": {
    "mode": "local",
    "controlUi": {
      "dangerouslyDisableDeviceAuth": true
    }
  },
  "agents": {
    "defaults": {
      "model": "claude-opus-4-5"
    }
  },
  "channels": {
    "telegram": {
      "enabled": true,
      "botToken": "YOUR_TELEGRAM_BOT_TOKEN",
      "dmPolicy": "open"
    }
  }
}
EOF

Replace YOUR_TELEGRAM_BOT_TOKEN with the token from @BotFather. The gateway.mode must be "local" — other values cause the gateway to exit immediately.

Step 3 — Run the container

Start OpenClaw with the config file mounted and your AI API key passed as an environment variable:

docker run -d \
  --name openclaw \
  --restart unless-stopped \
  --memory 2g \
  -p 8080:18789 \
  -v /opt/openclaw/config.json:/home/node/.openclaw/config.json \
  -e ANTHROPIC_API_KEY=sk-ant-YOUR_KEY \
  -e NODE_OPTIONS=--max-old-space-size=1536 \
  alpine/openclaw:latest \
  node /app/openclaw.mjs gateway

Step 4 — Verify OpenClaw is running

OpenClaw takes about 5 minutes to initialize on first start. You can watch the logs to confirm it's working:

# Follow container logs
docker logs -f openclaw

# Look for these lines indicating success:
# [gateway] WASM sandbox compiled
# [telegram] Polling started
# [gateway] Listening on 127.0.0.1:18789

Once you see the polling started message, open your Telegram app and send a message to your bot. You should get an AI response within a few seconds. Access the Control UI at http://YOUR_SERVER_IP:8080.

Method 3 — npm Install (For Developers)

Setup time: 15–30 minutes · Requires: Node.js 18+, npm, AI API keys

If you're a developer who wants to run OpenClaw directly on your machine or integrate it into an existing Node.js project, the npm install method gives you the most flexibility. This is also useful for local development and testing before deploying to a server.

Step 1 — Install OpenClaw globally

npm install -g openclaw

This installs the openclaw CLI globally. Verify the installation with:

openclaw --version

Step 2 — Initialize a config

Run the init command to generate a starter configuration file in the current directory:

openclaw init

This creates ~/.openclaw/config.json with sensible defaults. Open it in your editor and add your bot token and AI API key.

Step 3 — Edit the config file

The generated config will look similar to this. Fill in your credentials:

// ~/.openclaw/config.json
{
  "gateway": {
    "mode": "local",
    "controlUi": {
      "dangerouslyDisableDeviceAuth": true
    }
  },
  "agents": {
    "defaults": {
      "model": "claude-opus-4-5"
    }
  },
  "channels": {
    "telegram": {
      "enabled": true,
      "botToken": "YOUR_TELEGRAM_BOT_TOKEN",
      "dmPolicy": "open"
    }
  }
}

Step 4 — Start the gateway

# Set your AI API key
export ANTHROPIC_API_KEY=sk-ant-YOUR_KEY

# Start the gateway
openclaw gateway

The gateway starts and after ~5 minutes of initialization, your bot is live. The Control UI is available at http://localhost:18789. See the OpenClaw setup guide for more configuration options.

Running as a background service (Linux)

For a persistent installation on Linux, create a systemd service so OpenClaw restarts automatically:

# /etc/systemd/system/openclaw.service
[Unit]
Description=OpenClaw Gateway
After=network.target

[Service]
Type=simple
User=ubuntu
ExecStart=/usr/bin/openclaw gateway
Restart=always
RestartSec=10
Environment=ANTHROPIC_API_KEY=sk-ant-YOUR_KEY
Environment=NODE_OPTIONS=--max-old-space-size=1536

[Install]
WantedBy=multi-user.target

sudo systemctl daemon-reload
sudo systemctl enable openclaw
sudo systemctl start openclaw
sudo systemctl status openclaw

ClawMates vs Docker vs npm — Which Should You Use?

Here's a side-by-side comparison of all three OpenClaw installation methods:

For most users, ClawMates is the right choice. If you have specific self-hosting requirements or are integrating OpenClaw into an existing infrastructure, Docker or npm gives you full control. See the OpenClaw hosting overview for a deeper comparison.

OpenClaw Installation FAQ

More OpenClaw Resources