Key facts

  • Install OpenClaw on a VPS, connect an AI provider and Telegram bot, enable the background service, and troubleshoot the most common setup errors.
  • Operating system: a current Linux server such as Ubuntu or Debian. macOS and Windows are also supported through their documented installation paths.
  • Node.js: version 24.16+ or 26.1+; the project recommends Node 26.

OpenClaw is an open-source AI agent that runs on your own server. Unlike a browser chatbot, it can stay online around the clock, connect to Telegram, WhatsApp, and Slack, work with files, monitor websites, and run scheduled tasks.

Installing it still requires some technical preparation: a VPS, Node.js, SSH access, and provider credentials. An experienced administrator may finish the setup quickly, while a first-time user can easily get stuck on terminal commands or configuration errors.

This guide walks through a server installation and explains the main choices you will see during initial setup.

What OpenClaw is and what it can do

OpenClaw is a self-hosted AI assistant. This guide was reviewed against the project's current installation, Linux, dashboard, and Telegram documentation in October 2026.

FeatureTypical browser chatbotOpenClaw
AvailabilityWorks while the browser session is openRuns continuously on your server
MemoryUsually limited to the current service or sessionCan keep persistent agent context
ActionsMainly answers questionsCan use configured tools, files, and messaging channels
AccessWebsite or appTelegram, WhatsApp, Slack, and a web interface
AutomationUsually manual promptsScheduled tasks and automated actions

Common uses include a morning digest, price monitoring, email triage, and controlling Home Assistant through a messenger.

OpenClaw installation requirements

Practical server baseline

  • Operating system: a current Linux server such as Ubuntu or Debian. macOS and Windows are also supported through their documented installation paths.
  • Node.js: version 24.16+ or 26.1+; the project recommends Node 26.
  • Resources: size the server for the connected tools and workload, then monitor memory and disk use after launch.

What you need before starting

  • SSH access to the server.
  • An existing supported CLI login or an API key from an AI provider.
  • A Telegram account if you want to use the agent through Telegram.

Step-by-step server installation

Step 1. Connect over SSH

Open a terminal and connect to the server:

ssh root@your_server_ip

On Windows, use WSL2 or an SSH client. macOS and Linux already include an SSH client in the terminal.

Step 2. Install OpenClaw and a supported Node.js runtime

The official installer is the simplest supported path. It installs a compatible Node.js runtime when the server needs one:

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

After it finishes, confirm the installed versions:

node --version
openclaw --version

Step 3. Use npm only if you already manage Node.js

With npm 11.16 or newer, including npm 12, allow the OpenClaw package's installation script explicitly:

npm install -g openclaw@latest --allow-scripts=openclaw

On npm 11.15 or earlier, omit --allow-scripts=openclaw. Choose either the official installer or the manual npm route; you normally do not need to run both.

Step 4. Run onboarding and install the service

If the installer has not already opened onboarding, start the wizard and install the background service:

openclaw onboard --install-daemon

The wizard asks you to accept the security disclaimer, choose a setup mode and AI provider, enter an API key, select a default model, and configure a messaging channel.

Step 5. Create a Telegram bot

  1. Open Telegram and find the verified @BotFather account.
  2. Send /newbot.
  3. Choose a display name and a username ending in bot.
  4. Copy the token returned by BotFather and paste it into the OpenClaw setup wizard.

Treat the bot token like a password. Do not publish it or include it in screenshots.

Step 6. Approve pairing

Send a message to the new Telegram bot. If it returns a pairing code, approve that code on the server:

openclaw pairing approve telegram YOUR_CODE

Step 7. Verify the background service

Confirm that the gateway is running, then follow the logs while sending a harmless test message:

openclaw gateway status
openclaw logs --follow

# After leaving the live log with Ctrl+C, restart only if needed
openclaw gateway restart

Open the web dashboard

Start the dashboard with:

openclaw dashboard

The interface is normally available at http://127.0.0.1:18789/. For a remote server, create an SSH tunnel:

ssh -L 18789:localhost:18789 root@your_server_ip

Then open http://localhost:18789/ on your computer. The dashboard provides access to sessions, logs, channels, skills, and agent memory.

Choose an AI provider

OpenClaw can work with several providers. Availability, model names, and pricing change frequently, so check the current provider list during onboarding. Keep a fallback model from another provider if the agent performs important scheduled work.

Security precautions

  1. Use one trusted owner by default. A shared bot may expose conversation context or connected tools to other users.
  2. Assume prompt injection is possible. Do not give an experimental agent unrestricted access to critical systems.
  3. Protect tokens and API keys. Rotate them immediately if they are exposed.

Restrict the Telegram bot to your account, review logs regularly, and grant only the permissions each automation needs.

Common installation errors

Command not found: openclaw

The global npm binary directory may be missing from PATH:

export PATH="$PATH:$(npm config get prefix)/bin"

Add the line to ~/.bashrc if it fixes the command.

Pairing code expired

Send the bot another message and approve the new code.

Gateway not running

openclaw gateway start
# or
systemctl --user start openclaw-gateway

API key invalid

Check the provider dashboard, confirm that the key is active, and make sure the account has sufficient quota or balance.

A managed installation option

If you do not want to configure the VPS, terminal, service, and provider credentials yourself, you can request an OpenClaw installation. Confirm the current scope and price before ordering because server and AI-provider charges are normally separate.

Summary

  1. Prepare a supported server and connect over SSH.
  2. Install Node.js and OpenClaw.
  3. Run onboarding and connect an AI provider.
  4. Create a Telegram bot and approve pairing.
  5. Install the background service and verify the gateway logs.

After installation, test the agent with a harmless task before connecting email, files, calendars, or other sensitive systems.