Overview
OpenClaw is an open-source local AI assistant created by developer Peter Steinberger in November 2025. It runs on your own computer and interacts through various chat platforms like Telegram, WhatsApp, Discord, etc., enabling file operations, terminal command execution, browser control, and other automation tasks. By configuring APIYI service, you can get:🔒 Local Privacy
Data is processed locally, ensuring privacy and security
📱 Multi-Platform
Supports Telegram, WhatsApp, Discord, Slack, iMessage, and more
🛠️ Autonomous Execution
Not just chat - actually operates your computer and executes commands
🔌 Extensible
Supports custom skills and plugin extensions
Project Info: OpenClaw is an MIT-licensed open-source project. Website:
openclaw.ai, Repository: github.com/openclaw/openclaw.Prerequisites
System Requirements
- Node.js 22 or higher
- macOS / Linux / Windows
Install Node.js
- macOS (Homebrew)
- Other Systems
Install OpenClaw
Quick Configuration
OpenClaw uses a JSON configuration file located at~/.openclaw/openclaw.json.
Configure APIYI
Create or edit the configuration file~/.openclaw/openclaw.json:
Configuration Reference
| Field | Description |
|---|---|
agents.defaults.model.primary | Default model, format: provider/model-name |
models.providers | Custom model provider configuration |
baseUrl | API endpoint, use https://api.apiyi.com/v1 |
apiKey | Your APIYI key |
api | API type: openai-completions for OpenAI compatible, anthropic-messages for Anthropic compatible |
Run Setup Wizard
- Gateway configuration
- Chat channel connection (Web UI, Telegram, WhatsApp, etc.)
- Skills configuration
Start the Service
Usage Methods
Method 1: Web UI (Recommended)
The simplest method, no external services required:http://127.0.0.1:18789/ where you can chat directly in the web interface.
Method 2: Telegram Bot
- Search for
@BotFatheron Telegram - Send
/newbotto create a bot - Get the Bot Token
- Enter the Token during
openclaw onboard
Method 3: Other Platforms
OpenClaw also supports:- WhatsApp (scan QR to connect)
- Discord (requires creating a Bot)
- Slack, Signal, iMessage, Microsoft Teams, etc.
Core Skills
OpenClaw comes with rich built-in skills for various tasks:File Operations
| Skill | Function |
|---|---|
fs.read | Read files (text/images) |
fs.write | Write/create files |
fs.edit | Edit file content |
System Operations
| Skill | Function |
|---|---|
shell.exec | Execute terminal commands |
shell.process | Manage running commands |
browser.* | Browser automation (open pages, screenshot, click, etc.) |
Smart Features
| Skill | Function |
|---|---|
web_search | Web search |
web_fetch | Fetch web content |
memory_search | Search memory |
memory_get | Retrieve memory |
cron.* | Scheduled tasks (reminders, automation) |
tts | Text-to-speech |
Common Commands
Terminal Commands
| Command | Function |
|---|---|
openclaw onboard | Run setup wizard |
openclaw gateway start | Start Gateway service |
openclaw gateway restart | Restart Gateway service |
openclaw gateway stop | Stop Gateway service |
openclaw status | Check running status |
openclaw doctor | Diagnose configuration issues |
openclaw doctor --fix | Auto-fix configuration issues |
openclaw dashboard | Open Web control panel |
openclaw logs --follow | View real-time logs |
openclaw configure | Modify configuration |
openclaw update | Update to latest version |
Chat Commands
Commands available in the chat window:| Command | Function |
|---|---|
/help | Show help |
/new | Start new conversation |
/reset | Reset conversation |
/stop | Stop current task |
/think <level> | Set thinking depth |
/model <id> | Switch model |
/verbose on/off | Toggle verbose mode |
/status | Check status |
/skills | View available skills |
Usage Examples
Chat naturally with OpenClaw and it will understand and execute tasks:File Operations
Terminal Commands
Browser Control
Scheduled Tasks
Programming Assistance
Recommended Models
OpenClaw supports 200+ mainstream AI models through APIYI. Choose the right model for different tasks.View Model Recommendations
Check the latest scenario-based model recommendations, including text creation, programming, quick response, long document processing, and more.
Scenario-Based Recommendations
| Task Type | Recommended Model | Reason |
|---|---|---|
| Complex Task Execution | Claude Sonnet 4 | Strong comprehension, accurate execution |
| Daily Conversation | GPT-4.1 | Natural responses, good versatility |
| Code Writing | DeepSeek V3 | Strong coding ability, cost-effective |
| Long Document Processing | Gemini 2.5 Pro | Supports ultra-long context |
Troubleshooting
Can't connect to Telegram
Can't connect to Telegram
Telegram may require a proxy in some regions. Set up terminal proxy and restart Gateway:Or use Web UI directly (no proxy needed).
Configuration file format error
Configuration file format error
Run diagnostic command to auto-fix:
API connection failed
API connection failed
- Check if API key is correct
- Confirm
baseUrlis set tohttps://api.apiyi.com/v1 - Test connectivity:
How to view runtime logs
How to view runtime logs
How to update OpenClaw
How to update OpenClaw
How to add new chat channels
How to add new chat channels
Advanced Features
Multi-Model Configuration
Configure multiple models and switch between them using/model <id> in chat:
Custom Skills
OpenClaw supports custom skills. Create them in the~/.openclaw/workspace/skills/ directory.
Memory Feature
OpenClaw remembers your conversations and preferences. Use/memory to view and manage memory.