Core Concepts
Understanding OpenClaw's architecture and key concepts.
Overview
OpenClaw is built around several core concepts:
- Gateway: The central hub managing all communications
- Agents: AI entities that process and respond to messages
- Sessions: Conversation contexts
- Memory: Long-term knowledge storage
- Channels: Messaging platform integrations
Quick Links
| Concept | Description |
|---|---|
| Architecture | System architecture overview |
| Agent | How agents work |
| Agent Loop | Message processing cycle |
| Session | Conversation management |
| Memory | Long-term memory system |
| Models | AI model configuration |
How It Works
User Message → Channel → Gateway → Agent → Model → Response- User sends a message via a channel (WhatsApp, Telegram, etc.)
- Channel receives and forwards to the Gateway
- Gateway routes to the appropriate Agent
- Agent processes using the configured Model
- Response is sent back through the channel
Key Principles
- Local-first: All processing happens on your machine
- Privacy-focused: Your data stays with you
- Extensible: Add custom tools, skills, and channels
- Multi-platform: Works across different messaging apps