Quick Start Guide for OpenClaw (Lobster) Intelligent Automation Framework
This tutorial helps you get OpenClaw running quickly on your local machine or server, and connect it to an OpenAI Hub model API.
1. Install OpenClaw
Whether you’re using Mac, Linux, or Windows, it supports one-click installation. The script automatically detects Node.js (must be ≥ v22), and installs it for you if missing.
✅ macOS or Linux:
Open your terminal and run:
# Official installation script (recommended)
curl -fsSL https://openclaw.ai/install.sh | bash
# Use the domestic mirror script if your network is slow (more stable)
curl -fsSL https://open-claw.org.cn/install-cn.sh | bash
During installation, you'll be prompted for your computer password (input hidden, just press Enter afterward).
Then make the environment variables take effect:
# If you’re using zsh (default on Mac)
source ~/.zshrc
# If you’re using bash
source ~/.bash_profile
Verify installation success:
openclaw --version
If you see a version number (something like 2026.x.x), you’re good to go.
✅ Windows:
Open PowerShell as Administrator and execute:
iwr -useb https://openclaw.ai/install.ps1 | iex
Careful: you must “Run PowerShell as Administrator,” or the Node.js installation may hang.
2. Configure the Model API (using OpenAI Hub)
Once OpenClaw is installed, it launches a setup wizard asking which model you want to connect.
This is essentially telling it “where the brain is.”
Steps:
-
When the initialization screen appears, select
Yes→ press Enter. -
Enter “Quick Configuration (QuickStart).”
-
Choose “Custom API Access.”
-
When prompted for Base URL, input:
https://api.openai-hub.com/v1 -
Then input your API Key (retrievable from the openai-hub.com console):
sk-your-api-key -
Choose your model: You can select GPT-4, Claude-3, Gemini, DeepSeek, etc. All connect via OpenAI Hub, direct access in China—no proxy needed.
-
Press Enter to confirm configuration.
OpenClaw will test connectivity automatically; if you see “✅ Model connect success,” it means everything’s set.
3. Verify the Setup
In the terminal, run:
openclaw tui
This starts a terminal-based text UI, letting you chat with OpenClaw directly, e.g.:
Help me organize today’s to-do list.
If the model responds normally, it’s fully connected.
Common issue: if you get an Unauthorized error, check whether there are extra spaces or non-ASCII characters around your API Key.
4. Extend Skills (Optional)
OpenClaw Skills work like Chrome extensions—they’re its “hands and tools.”
If you want it to handle files, send emails, or monitor tasks, install Skill packs.
Install with:
# View available skills
openclaw skill list
# Install a skill, e.g. PDF processing
openclaw skill install pdf-kit
Once installed, you can trigger them with natural language, e.g.:
Compress this PDF to under 10MB.
By default, Skills are stored at:
~/.openclaw/skills
Do not change this directory—otherwise OpenClaw won’t find the files.
5. FAQ
Q1: Says Node.js not found?
Run node -v to check if it’s version 22 or higher. If not, reinstall.
Q2: Model configured but not responding?
Run:
openclaw service restart
Then try again.
Q3: Connection errors inside mainland China?
Use the OpenAI Hub domain https://api.openai-hub.com/v1 instead of the official one—direct access, no proxy required.
Next Steps
Try OpenClaw’s automation tasks—have it summarize daily reports on schedule, auto-organize files, or control your browser. It doesn’t just talk—it acts.