English Overview

中文文档

OpenClaw: Beginner to Expert

OpenClaw is an open-source, self-driven AI agent, created by Peter Steinberger. This book provides a comprehensive guide from getting started to production deployment, with deep dives into the underlying mechanisms and implementation principles.

OpenClaw Guide Cover

Highlights

Note: this English page is currently an overview. The full chapter-by-chapter content is maintained in Chinese; for complete reading, use the Chinese README and table of contents.

• Hands-on: Build a minimal working loop from scratch, with ready-to-use configuration templates

• Deep Dives: Detailed analysis of Gateway, Agent Loop, tool system, sessions and memory

• Production-Ready: Focus on reliability, security hardening, monitoring and troubleshooting

Target Audience & Prerequisites

• Target Audience: Individual users interested in AI agents, AI application developers, LLM engineers, system architects, etc.

• Prerequisites: Basic command-line experience is sufficient; programming knowledge is helpful but not required. For those unfamiliar with LLMs and AI agents, see AI Beginner Guide and Agentic AI Guide.

Book Structure

Part	Chapters	Overview
Part 1: Getting Started	Ch 1–4	Overview, setup, first conversation, configuration & model access
Part 2: Advanced Usage	Ch 5–8	Tools & skills, context memory, multi-agent collaboration, automation & ops
Part 3: Internals & Engineering	Ch 9–12	Gateway protocol, Agent Loop internals, reliability, plugin extensions
Part 4: Practice & Optimization	Ch 13–16	Case studies, performance & cost optimization, troubleshooting, AI ecosystem integration
Appendix	—	Glossary, config templates, troubleshooting checklist, API reference, command cheatsheet, version mapping, further reading

How to Read

Online

• GitBook Online

• Chinese Full Edition

Local Preview

This repository uses mdPress for building. For local preview:

brew tap yeasy/tap && brew install mdpress
mdpress serve

Other Markdown previewers can be used as auxiliary tools, but they are not the standard build chain for this repository.

5-Minute Quick Start

New to OpenClaw? Just three steps:

1. Install (1 min): curl -fsSL https://openclaw.ai/install.sh | bash

2. Initialize (2 min): openclaw onboard --install-daemon → follow the wizard to configure your model API key and install the background service

3. Chat (2 min): Run openclaw dashboard, type "hello" in the Control UI chat, receive an AI response — you're done! 🎉

See Chapter 2: Setup and Chapter 3: First Conversation.

Learning Paths

Different readers can choose their path:

Role	Core Chapters	Est. Time	What You'll Achieve
Individual User	1→2→3→5	2-3 hours	Build a personal WhatsApp/Telegram AI assistant
App Developer	1-7→12	8-10 hours	Develop custom tools, skills and multi-agent systems
Ops Engineer	2→3→8→11→14→15	6-8 hours	Production deployment, security hardening & troubleshooting
Architect	1→9→10→12→16	6-8 hours	Understand internals, design enterprise-grade agent architecture

Related Books

This book is part of an AI technology series:

Book	Relationship
AI Beginner Guide	Zero-to-one AI introduction
Prompt Engineering Guide	Agent prompt design fundamentals
Context Engineering Guide	Context management & memory architecture
Claude Guide	Claude MCP protocol, tools & Agentic Coding
Agentic AI Guide	General agent architecture & multi-agent patterns
AI Security Guide	Agent security design & attack/defense practices
LLM Internals	Deep dive into LLM architecture

Contributing

Welcome to submit Issues or PRs. Especially welcome: typo fixes, broken link repairs, case study additions, and reusable templates.

License

This book is licensed under CC BY 4.0.