OpenClaw Guide for Non-Programmers

A lot of friends have been asking me about OpenClaw lately, so I figured I’d write a short guide to answer the most common questions from people without a programming background.

Step 1: What Do I Need?

Find a spare computer. 
I’d recommend using an old computer dedicated to running it. 
Two reasons:

First, stability. You’ll probably want this assistant online 24/7 — handling important emails that come in at midnight, sending you a news digest every morning, and keeping an eye on things while you’re away. If you run it on your daily work laptop, it goes offline the moment you close the lid.

Second, security. This assistant will have access to your emails, calendar, and files. Giving it a separate machine, isolated from your work environment, means that even if something goes wrong with the configuration, it won’t affect your everyday setup. A five or six-year-old laptop is more than enough. Mac, Windows, or Linux — all work fine.

What if I don’t have a spare computer?

• Raspberry Pi — a few hundred RMB, extremely low power consumption, runs quietly in a corner
• Cloud server (VPS) — a few dozen to a hundred-something RMB per month, no hardware to worry about
• Mac mini — if you were planning to buy one anyway, this is the luxury option. One thing I strongly recommend: use a dedicated device, separate from your work computer.

Step 2: Does It Cost Money?

OpenClaw is open source. The cost comes from the AI model — think of it this way: you’ve hired an assistant, and the AI model is the assistant’s brain. My most straightforward advice: use the smartest, most capable model you can afford.

A dumb model means a dumb assistant. Why would you want a dumb assistant? On top of that, a dumb assistant is more easily manipulated by malicious instructions hidden in web pages. If it gets tricked once, the loss could be enough to cover a lifetime of model bills.

The capability gap between AI models is enormous. Cheap models can chat, but their comprehension, judgment, and reliability are noticeably weaker. You’ve already invested the time to set up this assistant — don’t cheap out on its brain. The time a smart assistant saves you will far exceed the model fees.

For light usage (a few conversations a day, occasional lookups), expect to pay roughly $5–$12 per month. Heavy usage will cost more. If you already have some AI subscriptions, ask a tech-savvy friend to take a look — you might be able to use them without any extra cost.

Step 3: Install It

The installation requires someone to type a few commands into the computer’s terminal. It’s not complicated, but if you’ve never used a terminal before, I’d suggest asking a tech-savvy friend to help. It takes about ten minutes. Once installed, OpenClaw comes with a built-in web chat interface — just open a browser and start talking to it.

But most people want to chat with it on their phone anytime, anywhere. You can connect it to a messaging app. I recommend Telegram (easiest to set up) or WhatsApp.

Step 4: Figure Out What It Can Do for You

Start with chatting.

When you first set it up, just treat it as a conversation partner. Ask it questions. Have it look things up, translate, draft emails, and organize your thoughts. Use it for a few days and get a feel for it.

Then gradually give it more permissions.

Once you feel confident it’s reliable, start granting more capabilities — reading your emails, managing your calendar, searching the web, running scheduled tasks. How do you grant permissions? Just ask it. Tell it “I want you to read my emails” or “I want you to send me a news digest every morning,” and it’ll tell you what configuration steps are needed. These steps are usually straightforward, and it’ll walk you through them.

It has memory.

Things you tell it, it remembers. Next time you chat, you don’t need to start from scratch. Every capability can be toggled on or off independently. Add them one by one. Build trust gradually.

Next Step: Set One Up for Your Family

No need to buy another computer. A single machine can run multiple independent AI assistants — one for each person. Each person’s assistant is completely separate:

• Different personalities — yours can be crisp and direct, your kid’s can be patient and gentle
• Separate memories — they can’t see each other’s chat history
• Different permissions — you can add restrictions to your kid’s assistant
• Different entry points — each person chats with their own bot

There’s another fun setup: create a group chat. Say you and your spouse create a group and add the assistant. You casually discuss “where should we eat this weekend” or “how should we plan next week’s schedule” in the group, and the assistant listens in the background. When you need it, @ it, and it can help look up restaurants, compare options, and note down decisions. When you don’t need it, it stays quiet — no unsolicited opinions.

The family group assistant is like an on-call butler: it’s there while you discuss things, and you just call on it when you need help.

What Can Family Members Use It For?

No tech skills required — if you can type in a messaging app, you’re good. Your spouse might use it like this:

• “Find some family-friendly activities for this weekend” — it searches, organizes, and lists options
• “Remind me to pick up the kids at 3 PM tomorrow” — it sends a message when the time comes
• “What does this English email say?” — forward it, and it translates and summarizes
• “Help me write a leave request email to the teacher” — tell it the reason, it drafts it, you confirm and send
• “We’re thinking about getting a new dishwasher, help me compare options” — it researches, makes comparisons, and gives recommendations
• Snap a photo of what’s in the fridge and ask “what should I cook tonight?” — it can look at the image and suggest recipes

Your kids might use it like this:

• “I don’t understand this math problem.” — Take a photo and send it, it’ll explain step by step
• “Check my English essay” — it’ll fix grammar and offer suggestions
• “Explain photosynthesis in simple terms” — more patient than a search engine, and you can keep asking follow-ups
• “I want to make a poster about space, give me some ideas” — it helps organize content and suggests layouts

How Is This Different from ChatGPT?

ChatGPT is a website. You open it, chat for a bit, close it, and next time it might have forgotten everything. OpenClaw runs on your own computer. It has persistent memory, can connect to your emails, calendar, and files, can act autonomously when you’re away, and is always reachable through the messaging apps you already use. One is a chat tool. The other is a personal assistant.

Finally: Is My Data Safe?

• All of OpenClaw’s data — chat history, memory, configuration — is stored on your own computer, not in the cloud.
• The content of your conversations with the AI model is sent to the model provider’s servers (e.g., Anthropic, OpenAI).