Documentation

Documentation

Welcome to OpenClaw documentation! Everything from getting started to advanced configuration.

Who are you? Where to start?

You are…Start with…
New to OpenClawQuick Start - Get up and running in 5 minutes
Ready to installInstallation - Platform-specific setup
Installed, want basicsCLI - Common commands reference
Want to connect appsChannels - WhatsApp, Telegram, etc.
Want to configure modelsProviders - Claude, OpenAI, local models
Want to customize agentAgent Tuning - Workspace, skills, memory
Ready for productionGateway - Remote access, security
Want to understand internalsDeep Dive - Call chains + key mechanisms

Core Concepts

Gateway

OpenClaw uses a gateway architecture where the Gateway process manages all channels and AI interactions. Supports both local and remote deployment.

Agent

The Agent is the core of the AI assistant, defined through workspace files (Identity, Soul, Tools, Skills, etc.) that specify capabilities, memory, and interaction style.

Channels

Channels are the entry points where OpenClaw interacts with users, supporting WhatsApp, Telegram, Discord, iMessage, and more.

Tools

Tools are the Agent’s capabilities for executing specific tasks, including browser control, file operations, system commands, and automation.

Providers

Providers configure AI models, supporting Anthropic Claude, OpenAI, local models (Ollama), and Chinese models.

Documentation Structure

Getting Started

Core Concepts

  • Concepts - Architecture, Gateway, Tools, Sandbox, Memory
  • Channels - Platform integration guides
  • Providers - Provider config, model selection, failover

Advanced

  • Agent Tuning - Workspace files, skills, memory, Heartbeat
  • Tools - Browser, exec, Web, automation
  • Platforms - Platform-specific notes
  • Deep Dive - Implementation-oriented reading tracks

Production

  • Gateway - Configuration, remote access, security, diagnostics

Reference

  • Reference - Configuration, HTTP API, templates, glossary

Get Help