Imagine messaging your phone and having an AI agent on your home computer respond — not just with an answer, but by actually doing something: searching the web, running a script, sending a follow-up message, all without you touching a keyboard. That is what OpenClaw makes possible, and it runs entirely on hardware you already own.
OpenClaw is an open-source, self-hosted AI agent that has become one of the most talked-about projects of 2026. If you have heard the name but are not sure where to start, this guide walks you through what it is, how to install it, and what you can realistically do with it as a complete beginner.
What Is OpenClaw?
OpenClaw is an autonomous AI agent that runs locally on your own machine — a laptop, a Mac Mini, a home server, or even a Raspberry Pi — and connects to messaging apps you already use, including WhatsApp, Telegram, Slack, and Discord. Unlike browser-based chatbots, it does not just respond to questions. It reasons about a request, picks the right tools, and executes actions: running scripts, reading files, automating workflows, sending messages back when the task is done.
The distinction that matters is the shift from “tools that can talk” to “assistants that can do,” as APIyi’s beginner guide puts it. You send a task through chat. It figures out the steps, executes them, and reports back.
OpenClaw is also completely open-source. You bring your own API key (more on that shortly), and the software itself is free. Your data stays on your hardware, not on a third-party cloud server.
What You Need Before You Install OpenClaw
Getting set up requires a few prerequisites. None of them are complicated, but skipping any one of them is the most common reason beginners get stuck.
Hardware
OpenClaw runs on Mac, Linux, and Windows (via WSL2). Many users dedicate a Mac Mini as their always-on OpenClaw machine — not because it is especially powerful, but because it runs 24/7 without tying up a primary computer, as noted in the Build MVP Fast guide. A standard modern laptop works fine for getting started.
Software requirements
You need Node.js version 22 or higher. This is the single most common cause of installation failure, according to Shareuhack’s setup tutorial. Run node -v in your terminal before anything else. If the version is below 22, update it first.
An Anthropic API key
OpenClaw connects to Claude models by default. You will need an Anthropic API key from console.anthropic.com. Anthropic charges per use — no subscription. Rough monthly estimates:
| Usage level | Estimated monthly cost |
|---|---|
| Light use (occasional tasks) | $3 – $15 |
| Typical non-developer | $20 – $60 |
| Heavy developer use | $200 – $1,000+ |
One thing worth understanding before you start: OpenClaw is an agent, not a chatbot. Each task triggers 5–10 API calls because the agent reads files, executes tools, reasons through steps, and revises its output — and every call re-sends the full conversation context. The costs add up faster than you would expect from a regular chatbot. Budget accordingly.
How to Install OpenClaw
The fastest installation method uses a single command. Open your terminal and run:
curl -fsSL https://openclaw.ai/install.sh | bash
Or via npm:
npm install -g openclaw@latest
If you see a “command not found” error after the npm install, your npm global bin directory is not in your system PATH. The AnotherWrapper installation guide covers this fix in detail.
Configure your API key
Once installed, OpenClaw will prompt you for your Anthropic API key on first launch. You can also set it as an environment variable:
export ANTHROPIC_API_KEY=your_key_here
One critical security step
After installation, change the Canvas Host binding to 127.0.0.1. By default, the Canvas interface may be accessible to other devices on your local network. Binding it to 127.0.0.1 restricts access to your machine only. Takes about 30 seconds and prevents unintended LAN exposure.
The whole process — from a clean machine to a running agent — takes roughly 15 minutes, according to Shareuhack’s documented walkthrough.
Understanding OpenClaw Skills
Skills are what make OpenClaw useful beyond answering questions. A Skill is a modular plugin — a SKILL.md description file plus one or more scripts — that extends what the agent can do. Install Skills to give OpenClaw new capabilities: web search, calendar access, GitHub integration, n8n workflow triggers, and more.
Skills are available through ClawHub, the community marketplace. Installing one is straightforward, but read the source code before you do. As 1Password’s security team has noted, the Skill format is portable across agent ecosystems, which makes it a real attack surface if you install untrusted packages without looking at them first. A bad Skill is not just broken software — it runs with the same permissions as the agent itself.
When you are starting out, stick to Skills with clear documentation, active maintainers, and community reviews. The built-in Skills that ship with OpenClaw are the safest place to begin.
Two practical use cases for beginners
Abstract capability descriptions are less useful than concrete examples. Here are two realistic scenarios for someone new to OpenClaw.
Use case 1: Personal research assistant via Telegram
You connect OpenClaw to your Telegram account. When you are out and want background on something — a company, a news story, a technical concept — you send a message to your OpenClaw bot. It searches the web, reads relevant pages, pulls out the key points, and sends you a summary back in Telegram. No browser, no desk required.
This is one of the most common beginner setups because it only requires the web search Skill and a Telegram bot token, both of which are free to configure.
Use case 2: Home automation trigger via WhatsApp
You connect OpenClaw to WhatsApp and install a home automation Skill linked to your smart home platform. Sending “turn off the living room lights” to your OpenClaw number triggers the action directly. More usefully, you can send compound instructions: “Turn off the lights, set the thermostat to 68, and remind me at 7am to check the washing machine.” OpenClaw handles each step in sequence and confirms when done.
This is what the Automate Your Life Tutorials YouTube channel describes as a shift from home automation to “life automation” — the agent handles compound tasks, not just single commands.
A note on security for new users
Running an autonomous agent on your home machine means it can execute real actions. That is the power of OpenClaw, and it is also where you need to pay attention.
Three things matter most for beginners. First, bind Canvas to 127.0.0.1 as described above. Second, review any Skill’s source code before installing from ClawHub — a malicious Skill runs with the same permissions as the agent. Third, be explicit about what your agent can and cannot do. OpenClaw supports permission scoping so you can restrict which directories it reads and which commands it can run.
The SlowMist OpenClaw Security Practice Guide on GitHub is a well-maintained community resource if you want to go further. For most home users, those three steps are enough to start safely.
Conclusion
OpenClaw is not a chatbot you visit in a browser tab. It is a self-hosted AI agent that lives on your own hardware, connects to the messaging apps you already use, and executes real tasks on its own. Getting it running takes about 15 minutes if your environment is ready — Node.js 22 or higher, an Anthropic API key, and one security configuration change after install.
Two things are worth getting right before you go deep: understand the token costs (agents burn through far more than chatbots do), and vet your Skills carefully before installing anything from ClawHub. Neither is hard to do. Both will save you from finding out the hard way.
Start with the one-line install script, connect it to Telegram, and try the web search Skill. That single setup will show you what an autonomous agent can actually do — and give you a foundation to build from there.
Frequently Asked Questions
Q: What is OpenClaw?
OpenClaw is an open-source, self-hosted AI agent that runs on your own hardware and connects to messaging platforms like WhatsApp, Telegram, Slack, and Discord. It receives requests through chat, reasons about them, and executes real actions — running scripts, reading files, automating workflows — rather than just responding with text.
Q: How do I install OpenClaw on my computer?
Run curl -fsSL https://openclaw.ai/install.sh | bash in your terminal, or use npm install -g openclaw@latest. Node.js version 22 or higher is required before you start — missing this is the most common reason the install fails. The full setup, including API key configuration, takes roughly 15 minutes.
Q: Does OpenClaw cost money to run?
OpenClaw itself is free and open-source. You pay for the AI model API calls — typically Anthropic’s Claude. Light users tend to spend $3–$15 per month. More regular non-developer use runs $20–$60 per month. Heavy use can reach several hundred dollars monthly because the agent makes 5–10 API calls per task, not just one.
Q: Is OpenClaw safe to run at home?
Yes, when configured correctly. Bind Canvas to 127.0.0.1, review Skill source code before installing anything from ClawHub, and scope the agent’s file and command permissions explicitly. The SlowMist Security Practice Guide is a reliable reference if you want a more thorough hardening checklist.
Q: What can OpenClaw actually do?
OpenClaw can search the web, read and write local files, run terminal commands, send and receive messages across multiple chat platforms, trigger automations in tools like n8n, and interact with services like GitHub and Slack through Skills. Its capabilities expand through modular plugins called Skills, which you install from the ClawHub marketplace.
Q: What hardware do I need to run OpenClaw?
OpenClaw runs on Mac, Linux, and Windows via WSL2. A modern laptop is sufficient for getting started. Many users dedicate a Mac Mini as a permanent always-on machine because it runs continuously without occupying a primary computer. A Raspberry Pi works too for lightweight tasks.