Most AI tools lock you into someone else’s platform. You pay monthly, your data lives on their servers, and you get whatever features they decide to ship.
OpenClaw works differently. This tutorial walks you through setting it up from scratch.
What Is OpenClaw?
OpenClaw is a free, open-source framework for building personal AI agents that run on your own machine. It powers Moltbot (formerly Clawdbot) — a self-hosted AI assistant you can talk to through Telegram, WhatsApp, Discord, or Signal.
What makes it worth setting up:
- You own it — Runs on your hardware. Your conversations and data never leave your machine
- Skills make it yours — Add modular abilities like image generation, YouTube summaries, or smart home control. Pick only what you need
- 3,500+ skills and growing — The community has built tools for media, productivity, development, finance, and more
- One agent, every channel — Connect once, message from any platform you already use
By the end of this guide, you’ll have a working AI agent on your messaging app of choice. The whole setup takes five steps.
What You Need Before Starting
Three things:
-
A computer running macOS, Linux, or Windows (WSL) OpenClaw runs on any Unix-compatible environment. Windows users need WSL2 installed.
-
A Claude API key OpenClaw uses Anthropic’s Claude as its AI backbone. Get your API key at console.anthropic.com.
-
A messaging channel (optional but recommended) Telegram, WhatsApp, Discord, or Signal. You can also use the Terminal for testing.
That’s it. OpenClaw is free and open-source. No cloud infrastructure, no Docker, no paid subscriptions.
Step 1: Install OpenClaw
Open your terminal and run:
curl -fsSL https://install.openclaw.dev | bash
This installs the OpenClaw runtime and the openclaw CLI tool. After installation, verify it worked:
openclaw --version
You should see a version number (e.g., openclaw 2.4.1).
Already Using Clawdbot?
If you’re migrating from Clawdbot, OpenClaw is the same project under a new name. Your existing configuration and skills carry over. Just update to the latest version:
openclaw update
Step 2: Configure Your API Key
Set your Claude API key:
openclaw config set api-key YOUR_API_KEY_HERE
Or add it to your environment:
export ANTHROPIC_API_KEY=your_key_here
To verify the key is working:
openclaw config check
Step 3: Connect a Messaging Channel
Pick a channel and follow the setup. Telegram is the easiest to start with.
Telegram
openclaw channel add telegram
This walks you through creating a Telegram bot via BotFather and linking it to your OpenClaw agent.
openclaw channel add whatsapp
Discord
openclaw channel add discord
Terminal (No Setup Needed)
For quick testing without configuring a messaging app:
openclaw chat
This opens a local chat interface in your terminal.
Step 4: Install Your First Skill
Skills are what make OpenClaw useful. They’re modular abilities you add to your agent.
Using ClawHub (Recommended)
ClawHub is the official marketplace. One command installs any skill:
clawhub install fal-ai
This installs fal-ai, a skill for generating images, videos, and audio through the fal.ai API.
Browse Skills First
Not sure what to install? Browse the full directory:
- Claw Directory — 400+ curated skills (from the 3,500+ in the ecosystem) with descriptions and categories
- Best Skills of 2026 — Editor’s picks across every category
Popular First Skills
| Skill | What It Does | Install Command |
|---|---|---|
| fal-ai | Image and video generation | clawhub install fal-ai |
| tube-summary | YouTube video summaries | clawhub install tube-summary |
| smart-ocr | Extract text from images | clawhub install smart-ocr |
| diagram-gen | Generate diagrams from code | clawhub install diagram-gen |
Verify Your Skill Is Loaded
After installing, check that it’s active:
openclaw skills list
Your new skill should appear in the list. Test it by sending a message:
/fal-ai generate a photo of a sunset over mountains
Step 5: Start Your Agent
Launch Moltbot:
openclaw start
Your agent is now running and connected to your messaging channels. Send it a message to test.
To run it in the background:
openclaw start --daemon
Understanding the Skill System
Understanding how skills work helps you get more out of OpenClaw — and makes building your own skills straightforward.
Where Skills Live
All skills are stored in:
~/.openclaw/skills/
├── fal-ai/
│ └── SKILL.md
├── tube-summary/
│ └── SKILL.md
└── smart-ocr/
└── SKILL.md
Each skill is a folder containing a single SKILL.md file — plain Markdown that tells the agent what the skill does and when to use it. This means you can read, edit, or write skills without touching any code.
Skill Categories
Skills fall into 10 categories:
- Media — Image generation, video editing, audio processing
- Productivity — Task management, note-taking, scheduling
- Development — Code tools, API design, debugging
- Communication — Email, messaging, notifications
- Automation — Workflows, integrations, triggers
- AI Tools — LLM utilities, prompt engineering, model access
- Data — Analysis, visualization, database tools
- Smart Home — IoT control, device management
- Finance — Budgeting, trading, crypto tools
- Trellis — Workflow systems and templates
Manual Installation
For skills not on ClawHub, or your own custom skills:
mkdir -p ~/.openclaw/skills/my-skill
Create ~/.openclaw/skills/my-skill/SKILL.md:
# My Custom Skill
> Brief description of what this skill does.
## Usage
Describe how to use this skill...
## Examples
- "Do X with this skill"
- "Perform Y action"
Then reload:
openclaw skills reload
What to Do Next
You now have a working AI agent connected to your messaging app, with a skill installed and running. That puts you ahead of most people who are still reading about AI agents instead of using one.
Find the Best Skills
Not sure which skills to install first? Read our Best OpenClaw Skills in 2026 guide — we tested hundreds and picked the top ones across every category.
Browse by Category
Browse by category on Claw Directory:
- Media Skills — If you work with images or video
- Development Skills — If you’re a developer
- Productivity Skills — If you want a personal assistant
Build Your Own Skill
Creating a skill is just writing a Markdown file. The SKILL.md format is straightforward — no programming required for basic skills.
Join the Community
- OpenClaw GitHub — Source code and issue tracker
- Official Docs — Full documentation
- Ecosystem Sites — 50+ community resources, tools, and guides
Secure Your Setup
Running a self-hosted AI agent means taking security seriously. Check the security articles on Claw Directory for best practices on API key management, network isolation, and skill auditing.
FAQ and Troubleshooting
Is OpenClaw free?
Yes. OpenClaw is free and open-source under the MIT license. You only pay for the Claude API usage (Anthropic charges per token). Most personal use costs a few dollars per month.
Does OpenClaw work on Windows?
Yes, through WSL2 (Windows Subsystem for Linux). Install WSL2 first, then follow the same steps as Linux users.
What’s the difference between OpenClaw, Moltbot, and Clawdbot?
OpenClaw is the framework. Moltbot is the AI agent built on it (formerly called Clawdbot). Think of OpenClaw as the engine and Moltbot as the car.
”Command not found: openclaw”
Your shell didn’t pick up the new PATH. Run:
source ~/.bashrc # or ~/.zshrc for macOS
“API key invalid”
Double-check your key at console.anthropic.com. Make sure you copied the full key without extra spaces.
”Skill not loading”
Check that:
- The SKILL.md file is in the correct directory (
~/.openclaw/skills/skill-name/) - The markdown format is valid
- You ran
openclaw skills reloadafter adding it
”Permission errors on skills”
Some skills need their own API keys (like fal-ai needs a fal.ai key). Check the skill’s documentation for required environment variables.
Summary
| Step | What You Did |
|---|---|
| 1 | Installed OpenClaw via the install script |
| 2 | Configured your Claude API key |
| 3 | Connected a messaging channel (Telegram, etc.) |
| 4 | Installed your first skill from ClawHub |
| 5 | Started your Moltbot agent |
OpenClaw gives you a personal AI agent that runs on your machine, connects to your messaging apps, and does what you tell it to — through skills you choose.
Start with one skill. Use it for a few days. Add another when you find something you wish your agent could do. That’s how the best setups get built.