OpenClaw

OpenClaw is a powerful, open-source AI agent platform that turns a standard Large Language Model (like Claude or GPT-4) into an autonomous assistant that lives on your hardware. Unlike a typical chatbot, it can “do” things: read your files, manage your calendar, execute code, and browse the web—all from within your favorite messaging apps.
Author

Benedict Thekkel

1. Everything to Know About OpenClaw

OpenClaw (formerly known as Clawdbot or Moltbot) is designed for users who want the power of an “Agentic” AI without sacrificing privacy or local system access.

Core Architecture

The system operates on a Hub-and-Spoke model: * The Gateway: A central Node.js process that handles routing, security, and connections. * The Brain (LLM): You provide an API key (Anthropic, OpenAI, or a local model via Ollama). * The Body (Tools): Local skills that allow the AI to interact with your filesystem, shell, and browser. * Channels: Integrations with WhatsApp, Telegram, Discord, Slack, and iMessage.

Key Features

  • Persistent Memory: It remembers context across days and different apps. If you tell it something on WhatsApp, it knows it when you talk to it on Slack later.
  • Action-Oriented: It doesn’t just talk; it executes. You can ask it to “Summarize my last 5 emails and create a task list in my todo.md file.”
  • Data Privacy: Your conversation history and local files stay on your machine. Only the specific prompts required for “reasoning” are sent to the LLM provider.
  • Markdown-Based: Configuration and memory are stored as plain text/Markdown files, making them easy for humans to read and audit.

2. How to Set Up OpenClaw

Setting up OpenClaw takes about 10–15 minutes. It is best installed on a machine that stays on (like a home server, Mac Mini, or a VPS), but it works fine on a standard laptop.

Prerequisites

  • Node.js: Version 22 or later is required (Node 24 is recommended).
  • API Key: An Anthropic (Claude), OpenAI, or OpenRouter key.
  • Terminal Access: Basic comfort with command-line interface (CLI).

Step-by-Step Installation

Step 1: Install the CLI

Open your terminal and run the global installation command:

npm install -g openclaw@latest

Note: On Linux or macOS, you may need to prefix this with sudo if you hit permission errors.

Step 2: Run the Onboarding Wizard

The onboarding command handles the initial configuration, including setting up the background service (daemon).

openclaw onboard --install-daemon

During this process, the wizard will ask you for: 1. Your LLM Provider: Choose between Anthropic, OpenAI, etc. 2. Your API Key: Paste your key here. 3. Gateway Setup: It will default to port 18789.

Step 3: Verify the Gateway

Check if the system is running correctly:

openclaw gateway status

If it’s live, you can open the browser-based dashboard to start chatting immediately:

openclaw dashboard

Step 4: Connect a Channel (e.g., Telegram)

To use OpenClaw from your phone, Telegram is the easiest bridge: 1. Message @BotFather on Telegram and use /newbot to create a bot. 2. Copy the API Token they give you. 3. In your terminal, run: bash openclaw channel connect telegram --token "YOUR_TELEGRAM_TOKEN" 4. Send a message to your new bot, and OpenClaw will reply!

🛡️ Critical Safety Tips

  • Sandbox Your Environment: It is highly recommended to run OpenClaw in a virtual machine (VM) or a dedicated Docker container rather than your “daily driver” computer, as the agent has the power to run shell commands.
  • Human-in-the-Loop: By default, you should keep “Human Approval” on for sensitive actions like deleting files or sending emails.
  • Set Spend Limits: Always set a hard billing limit on your Anthropic/OpenAI dashboard to prevent an accidental “infinite loop” from draining your account.
Back to top