Skip to content
Β 
Β 

Repository files navigation

🦞 LiteClaw

Lightweight Light Weight AI Assistant

LiteClaw is a lite weight AI Assistant gateway that connects LLMs to your world through messaging platforms (WhatsApp, Telegram, Slack). It enables powerful capabilities like Vision-powered OS control (mouse/keyboard) for browsers and desktop apps, shell execution, and proactive task management.

LiteClaw Logo

Note: LiteClaw is a lightweight, Python-based version of OpenClaw, focusing on speed, modularity, and easy deployment while providing the core agentic capabilities.

πŸ“Έ Demos

Just like its powerful parent project OpenClaw, LiteClaw (the lite mini version) captures user commands and executes them precisely, providing real-time feedback and screenshots.

Printed Demo Screenshot Demo

License: MIT Python 3.12+

✨ Features

  • πŸ€– Multi-LLM Support: Works with OpenAI, OpenRouter, Groq, DeepSeek, and local models (Ollama)
  • πŸ‘οΈ Vision OS Control: Cross-platform control (Windows, Linux, MacOS) via mouse movements and keyboard commands (Powered by Gemini 3 Flash and other Vision models)
  • πŸ“‘ Multi-Channel Support: WhatsApp, Telegram, Slack
  • πŸ’“ Proactive Heartbeat: Define periodic autonomous tasks in HEARTBEAT.md
  • 🧠 Adaptive Personality: Soul/Personality/Subconscious memory that evolves over time
  • πŸ’‘ Autonomous Innovation: A built-in "Subconscious" system that triggers random technical experiments and self-improvement tasks
  • πŸ”§ Shell Execution: Run terminal commands safely
  • πŸ“… Cron Jobs: Schedule recurring tasks via API
  • 🧡 Sub-Agents: Delegate long-running tasks to background agents

πŸš€ Quick Start

1. Install LiteClaw

# Clone the repository
git clone https://github.com/Pr0fe5s0r/LiteClaw.git
cd LiteClaw

# Create virtual environment
python -m venv .venv
source .venv/bin/activate  # Linux/Mac
# .venv\Scripts\activate   # Windows

# Install in development mode
pip install -e .

2. Run Onboarding Wizard

liteclaw onboard

This interactive wizard will:

  • Set up your work directory (~/liteclaw on Mac/Linux, C:\liteclaw on Windows)
  • Configure your LLM provider and API key
  • Set up WhatsApp bridge (optional)
  • Pair WhatsApp via QR code

3. Start the Agent

liteclaw run

πŸ“– CLI Commands

Command Description
liteclaw onboard Run the setup wizard
liteclaw run Start the gateway (FastAPI + Node Bridge)
liteclaw run --no-bridge Start without the Node.js bridge
liteclaw config View current configuration
liteclaw status Check system status

βš™οΈ Configuration

After onboarding, your config is stored in:

  • Mac/Linux: ~/liteclaw/config.json
  • Windows: C:\liteclaw\config.json

Supported LLM Providers

Provider Base URL
OpenAI https://api.openai.com/v1
OpenRouter https://openrouter.ai/api/v1
Groq https://api.groq.com/openai/v1
DeepSeek https://api.deepseek.com/v1
Ollama (Local) http://localhost:11434/v1

WhatsApp Setup

  1. During onboarding, select "WhatsApp (requires phone scan)"
  2. Scan the QR code with your WhatsApp mobile app
  3. Messages from your allowed numbers will be processed by the agent

Note: Add your WhatsApp ID to WHATSAPP_ALLOWED_NUMBERS in config.json to filter who can interact with the bot.

πŸ—οΈ Architecture

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”     β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”     β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚   WhatsApp      │────▢│   Node Bridge    │────▢│   FastAPI   β”‚
β”‚   (Phone)       │◀────│   (port 3040)    │◀────│  (port 8009)β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜     β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜     β””β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”˜
                                                        β”‚
                                                        β–Ό
                                                 β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
                                                 β”‚  LLM Agent  β”‚
                                                 β”‚  (LiteLLM)  β”‚
                                                 β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

πŸ“‚ Project Structure

LiteClaw/
β”œβ”€β”€ src/liteclaw/
β”‚   β”œβ”€β”€ agent.py         # Main AI agent logic
β”‚   β”œβ”€β”€ cli.py           # CLI commands
β”‚   β”œβ”€β”€ config.py        # Settings management
β”‚   β”œβ”€β”€ main.py          # FastAPI application
β”‚   β”œβ”€β”€ memory.py        # Conversation memory
β”‚   β”œβ”€β”€ scheduler.py     # Cron job manager
β”‚   β”œβ”€β”€ tools.py         # Agent tools (shell, browser, etc.)
β”‚   β”œβ”€β”€ AGENT.md         # Agent personality/instructions
β”‚   β”œβ”€β”€ HEARTBEAT.md     # Proactive task definitions
β”‚   β”œβ”€β”€ SOUL.md          # Long-term memory
β”‚   └── bridge/          # Node.js WhatsApp bridge
β”œβ”€β”€ config.json          # Your configuration
β”œβ”€β”€ requirements.txt
└── pyproject.toml

⚠️ Disclaimer & Status

IMPORTANT: LiteClaw is currently in active testing/beta mode.

Security Warning: This application currently has no built-in sandboxing or advanced security policies for tool execution (like shell or browser). It should be used with extreme caution.

  • Do not run this on a production server.
  • Avoid giving the agent access to sensitive data or high-privilege environments.
  • You are responsible for the actions the agent takes on your behalf.

🀝 Contributing

LiteClaw is an evolving project and we are looking for contributors and new ideas! Whether it's fixing bugs, adding new bridge integrations (Discord, Matrix, etc.), improving the core agent logic, or suggesting creative new use cases, your input is welcome.

Please see our Contributing Guide for more details on how to get involved, join discussions, and submit code.

πŸ› οΈ Development

# Install dev dependencies
pip install -e ".[dev]"

# Run with hot reload
liteclaw run

πŸ“„ License

MIT License - see LICENSE for details.

πŸ™ Credits

About

This is a light-weight version of OpenClaw project.

Resources

Contributing

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages