LiteClaw is a high-performance personal AI assistant written in Golang. It is designed to be resource-efficient, fast, and easily deployable as a single binary.

• Single Binary: No complex runtime dependencies (like Node.js, Python envs). Just download and run.
• Low Footprint: Extremely low memory usage (~10MB idle) and instant startup time (< 1s).
• Cross-Platform: Runs on macOS, Linux, and Windows.

• Multi-Provider Support: Native support for top-tier LLMs including:
  • OpenAI (GPT-4o, o1)
  • Anthropic (Claude 3.5 Sonnet)
  • Google Gemini
  • DeepSeek (V3, R1)
  • Alibaba Qwen
  • Minimax
  • Moonshot (Kimi) / Zhipu (GLM)
  • Ollama (Local models) & OpenRouter
• Agent Runtime: Supports tool calling, streaming responses, and long-running sessions.

Seamlessly connect your AI agent to your favorite chat platforms:

• Global: Telegram, Discord
• China-Specific: QQ (腾讯QQ), Feishu (飞书), DingTalk (钉钉), WeCom (企业微信)
• Apple: iMessage (macOS only)
• Experimental/Planned: Slack, WhatsApp, Signal, Matrix

Empower your agent with built-in capabilities:

• File System: Read, write, and edit files.
• Shell Execution: Run terminal commands safely.
• Browser Automation: Control web browsers for automation tasks (via Chrome/Playwright).
• Web Search: Access real-time information via Brave Search (MCP).
• Process Management: Manage system processes.
• Memory: Persistent note-taking and context retention.

• Model Context Protocol (MCP): Full support for the MCP standard, allowing connection to any MCP-compatible server for unlimited tool extensions.
• Skill System: Define new agent capabilities using simple Markdown files (SKILL.md).

• Cron Jobs: Built-in scheduler for recurring tasks (e.g., "Summarize news every morning at 8 AM").
• Natural Language Scheduling: Support for "every 24h" or cron expressions.

LiteClaw/
├── cmd/liteclaw/          # Main entry point
├── configs/               # Example configs (example.liteclaw*.json)
├── extensions/            # Channel adapters (Telegram, Discord, QQ, etc.)
├── internal/              # Core logic
│   ├── agent/             # AI Agent (Tools, Prompts, Skills)
│   ├── gateway/           # HTTP/WS server
│   ├── browser/           # CDP/Browser automation
│   ├── cron/              # Scheduler
│   ├── pairing/           # DM security
│   └── cli/               # CLI commands
├── mcp/                   # Model Context Protocol client
└── skills/                # Built-in skills library

• Go 1.24+
• make (optional, for easy building)

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

# Build the binary
make build
# or: go build -o liteclaw ./cmd/liteclaw

# Verify installation
./liteclaw version

# Run the onboarding wizard
./liteclaw onboard

The wizard will guide you through:

1. Model selection (provider + model)
2. API key configuration
3. Workspace directory
4. Channel setup (optional)

LiteClaw reads config from:

• ~/.liteclaw/liteclaw.json — Main configuration
• ~/.liteclaw/liteclaw.extras.json — MCP servers and extensions

Example configs are available in the configs/ directory.

# Foreground mode
./liteclaw gateway start

# Background (detached) mode
./liteclaw gateway start --detached

# Check status
./liteclaw status

# Stop gateway
./liteclaw gateway stop

• Control UI (React/Vue Web Dashboard)
• WebChat interface
• Docker support & Dockerfile
• Tailscale Serve/Funnel integration
• Doctor diagnostics command
• Full WhatsApp/Signal support (Baileys/signal-cli bridges)
• OAuth authentication for Claude/ChatGPT Pro subscriptions

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Run go test ./... before submitting
4. Submit a Pull Request

MIT License. See LICENSE for details.

LiteClaw is inspired by OpenClaw. Special thanks to the OpenClaw community for the original design concepts.