OpenClaw is one of the most powerful AI assistant platforms available today — but getting it running from scratch takes some technical know-how. This guide walks you through what's involved, what you need, and when it makes sense to just hire someone to do it for you.
What Is OpenClaw?
OpenClaw is software that turns your device into a personal AI assistant. Unlike ChatGPT (which runs in a browser), OpenClaw runs on your hardware and connects to your messaging apps — WhatsApp, Telegram, and others. Your conversations stay on your device, and the AI is configured specifically for how you work.
It's popular with Malaysian SME owners who want an AI assistant for customer replies, admin tasks, and daily operations — without paying a monthly subscription forever.
What You Need Before Starting
Before any setup begins, you'll need:
- A compatible device — Mac mini (M-series recommended), a Windows PC with 16GB+ RAM, or a VPS
- Node.js installed — OpenClaw is a Node.js application
- An AI model — either a local model via Ollama, or an API key from Anthropic or OpenAI
- A messaging channel — WhatsApp Business API access or a Telegram bot token
- 30–90 minutes of uninterrupted time — the first setup always takes longer than expected
If that list already looks unfamiliar, that's okay — it's a sign that a done-for-you setup might save you a lot of frustration.
The Setup Process — Step by Step
Step 1: Install Node.js and Dependencies
OpenClaw requires Node.js (version 18 or higher). On a Mac, this usually means using Homebrew:
brew install node
On Windows, you'll download the installer from nodejs.org. On Linux/VPS, use nvm to manage Node versions cleanly.
Step 2: Install OpenClaw
Once Node.js is ready, you install OpenClaw via npm:
npm install -g openclaw
This downloads the core package. The first install includes running openclaw setup to go through the initial configuration wizard.
Step 3: Configure Your AI Model
This is where most people hit their first roadblock. You have two options:
Option A: Local model via Ollama (recommended for privacy)
- Install Ollama (
brew install ollamaon Mac) - Pull a model:
ollama pull llama3.2orollama pull qwen2.5 - Configure OpenClaw to use
http://localhost:11434as the model endpoint
Option B: Cloud API (simpler, but ongoing cost)
- Get an API key from Anthropic (for Claude) or OpenAI (for GPT)
- Add the key to your OpenClaw config file
- Cloud models are generally smarter but your data leaves your device
For most Malaysian SME owners, we recommend starting with a local model — it keeps your business data private and there's no per-message cost.
Step 4: Connect Your Messaging Channel
This is where things get specific to how you want to use OpenClaw.
For WhatsApp: You'll need access to the WhatsApp Business API. This isn't the same as the WhatsApp Business app you download from the App Store — it's a separate API integration that requires either a Meta developer account or a third-party connector.
For Telegram: Easier to set up. Create a bot via BotFather (search @BotFather on Telegram), get a bot token, and configure it in OpenClaw.
Step 5: Write Your System Prompts
This is the most important step — and the one most tutorials skip over.
System prompts are instructions that tell your AI assistant who it is, what it knows about your business, and how to respond. A bad prompt gives you a generic assistant. A good prompt gives you an assistant that sounds like it actually works for your company.
For example, a restaurant owner's system prompt might include:
- Your menu items and prices
- Your opening hours and location
- How to handle reservation requests
- Common questions customers ask
Getting this right takes iteration. Expect to spend 1–2 hours writing and testing prompts before you're happy with the output.
Step 6: Start the Gateway and Test
OpenClaw runs a local gateway server that connects everything together. Start it with:
openclaw gateway start
Then send a message to your configured channel and see if the assistant responds correctly. If it doesn't, you're debugging connection configs, port issues, or model errors — which is normal for a first setup.
How Long Does It Take?
Realistically:
- If you're technical and follow every step correctly: 2–4 hours
- If you hit errors (normal): 4–8 hours
- If you've never used a command line before: Much longer, or possibly not at all
The setup is documented and doable — but it assumes comfort with terminal commands and basic networking concepts.
When to Use a Professional Setup Service
If any of the following sounds like you, it's worth considering a done-for-you setup:
- You don't want to spend hours reading technical docs
- You want the AI configured for your specific business from day one, not a generic chatbot
- You've tried following setup guides and hit errors you couldn't fix
- Your time is worth more than the RM5,000 setup fee
A professional setup gets you a working, tested, customised OpenClaw assistant in 48 hours — with a training session and 30 days of support included.
OpenClaw Setup Services in Malaysia
OpenClawMY offers done-for-you OpenClaw setup for Malaysian businesses. We handle everything from installation to custom prompts to channel integration — in person (KL, Penang, JB) or remotely anywhere in Malaysia.
RM5,000 flat fee. No subscriptions. 30-day support included.
Setup options:
- Mac mini setup — recommended hardware
- Windows PC setup — if you already have a Windows machine
- VPS setup — for always-on, no local hardware required
Book a discovery call if you're ready to get started or want to ask a question before committing.