Skip to content

Introduction

This course is a deep technical dive into how OpenClaw works internally — written for developers and power users who want to understand the system beyond what the official documentation covers.

What this is not:

  • A getting-started guide (see docs.openclaw.ai for that)
  • A guide to model selection or the TUI
  • A tutorial on connecting channels

What this is:

  • An architectural deep dive into every major subsystem
  • Explanations of how things fit together internally
  • Source-backed analysis of real implementation decisions
  • A reference you can return to when building on top of OpenClaw
#ChapterWhat you’ll learn
1Gateway ArchitectureThe WebSocket broker, routing, agent dispatch
2Workspace StructureFile hierarchy, loading order, precedence
3The Agent LoopTurn lifecycle, tool execution, thinking levels
4Session ManagementKey construction, isolation, compaction, storage
5Channel & Account ModelHow channels work, multi-account, plugin interface
6Multi-Agent RoutingBindings, isolation, cross-agent communication
7Memory SystemFile-based memory, context injection, search
8Skills & ToolsBuilt-in tools, AgentSkills spec, skill loading
9AutomationCron, hooks, standing orders, TaskFlow
10Security & PermissionsExec policy, sandboxing, auth profiles

Built by LangCraft. Source analysis based on OpenClaw v2026.4.x.