Chapter 1: What is OpenClaw?

OpenClaw is an open-source, self-hosted AI assistant gateway that bridges your favorite messaging apps — WhatsApp, Telegram, Slack, Discord, iMessage, Signal, Microsoft Teams, and more — directly to powerful AI coding agents. Instead of opening a terminal or a browser, you simply send a message from wherever you already are, and OpenClaw's AI agents respond, write code, run tasks, and return results — all in the same chat window.

This chapter gives you a complete picture of what OpenClaw is, why it exists, how it differs from other AI tools, and what its core concepts mean.

The Problem OpenClaw Solves

Modern AI coding tools like Claude Code are powerful, but they live inside a terminal or a desktop app. To use them you need:

A laptop or desktop
A terminal window open
Knowledge of CLI commands
Direct access to your codebase

This works well for solo developers at their desks. But real teams have:

Developers on mobile who need quick answers
Non-technical teammates who want AI help without learning a CLI
Multiple messaging platforms already in use (Slack for team, WhatsApp for clients)
A need for AI that persists across conversations and knows the project context

OpenClaw was built to solve exactly this. It acts as a gateway — a central server that sits between your team's messaging apps and your AI agents — so anyone on any device can access the full power of AI coding assistance.

What OpenClaw Actually Is

OpenClaw is three things at once:

Role	Description
AI Gateway	Routes messages from any channel to the right AI agent
Multi-Channel Hub	Connects WhatsApp, Telegram, Slack, Discord, Signal, iMessage, Teams, Matrix, and more
Self-Hosted Platform	Runs on your own server — your data never leaves your infrastructure

Think of it like a smart telephone exchange from the old days — but instead of routing phone calls, it routes AI conversations. Every message that comes in gets picked up, sent to the right AI brain, and the response is delivered back to the original sender.

OpenClaw vs. Claude Code

Students often ask: "I already know Claude Code — why do I need OpenClaw?" The answer is that they solve different problems.

Feature	Claude Code	OpenClaw
Interface	Terminal (CLI)	Any messaging app
Access	Local machine only	Any device, anywhere
Users	Individual developer	Entire team
Setup	Install CLI, open terminal	Deploy gateway, connect channels
AI Models	Claude only	Claude, GPT, Gemini, and more
Skills/Plugins	Built-in tools	ClawHub marketplace
Data Control	Local files	Self-hosted server
Mobile support	No	Yes (via WhatsApp, Telegram, etc.)

Claude Code is a precision tool for a developer at a keyboard. OpenClaw is a platform for a team spread across devices and time zones.

Core Concepts

Before you install anything, understand these six building blocks of OpenClaw. Every other chapter in this book builds on them.

1. Gateway

The Gateway is the heart of OpenClaw — the central Node.js server you install on your machine or cloud host. It:

Listens for incoming messages from all connected channels
Authenticates users and enforces workspace permissions
Routes messages to the correct agent
Returns responses back through the correct channel

You start the gateway with:

openclaw start

When the gateway is running, all your connected channels are live.

2. Channels

A Channel is any messaging platform connected to your OpenClaw gateway. Out of the box, OpenClaw supports:

WhatsApp (via WhatsApp Business API or Baileys)
Telegram (via Bot API)
Slack (via Slack App)
Discord (via Discord Bot)
iMessage (via macOS node)
Signal (via signal-cli)
Microsoft Teams
Google Chat
Matrix / Element

Each channel has its own configuration inside ~/.openclaw/openclaw.json. You can have multiple channels active at the same time — your team might use Slack while your clients use WhatsApp, and both connect to the same OpenClaw gateway.

3. Agents

An Agent is the AI brain behind OpenClaw. When a user sends a message, the gateway picks the right agent and forwards the message to it. The agent processes the request and returns a response.

OpenClaw supports multiple agent backends:

Claude (Anthropic) — default, recommended for coding tasks
GPT-4 / GPT-4o (OpenAI)
Gemini (Google)
Local models via Ollama

You configure your agents in openclaw.json and can assign different agents to different workspaces or channels.

4. Skills

Skills are capabilities you add to your agents — similar to plugins or tools. A default OpenClaw installation includes skills for:

Reading and writing files
Running shell commands
Browsing the web
Scheduling tasks (cron jobs)
Running code in a sandbox

You can install additional skills from ClawHub, the official skill registry, or build your own. Skills make OpenClaw extensible — your agents can do exactly what your workflow requires.

5. Workspace

A Workspace is a logical grouping of users, channels, and agents. Think of it like a Slack workspace or a GitHub organization — it defines who has access to what.

Examples:

A frontend-team workspace with Slack and specific Claude agent
A client-support workspace with WhatsApp and a GPT-4 agent
A personal workspace with Telegram and full tool access

Workspaces let you keep concerns separated and permissions controlled.

6. ClawHub

ClawHub is the official marketplace for OpenClaw skills, agent templates, and integrations. Available at openclawlab.com, it lets you:

Browse community-built skills
Install skills with a single command
Publish your own skills for others to use

openclaw skill install <skill-name>

ClawHub makes OpenClaw a living ecosystem, not just a fixed tool.

How a Message Flows Through OpenClaw

Here is what happens from the moment you send a message to the moment you get a reply:

You (WhatsApp) → Channel Adapter → Gateway Router → Agent (Claude)
                                                          ↓
You (WhatsApp) ← Channel Adapter ← Gateway Router ← Response

Step by step:

You type a message in WhatsApp: "Refactor this function to use async/await"
The WhatsApp channel adapter receives the message and passes it to the gateway
The gateway checks your workspace, identifies the assigned agent, and attaches relevant skills
The agent (Claude) processes your message with the codebase context
Claude returns the refactored code
The gateway sends the response back through the WhatsApp adapter
You see the refactored code in your WhatsApp chat

The entire round trip typically takes 3–10 seconds depending on the task complexity and the AI model used.

Why Self-Hosted?

OpenClaw is MIT licensed and fully self-hosted. This matters because:

Privacy: Your code, prompts, and responses never pass through a third-party server
Cost control: You pay only your AI API costs — no OpenClaw subscription fees
Customization: You can modify the source, add custom channels, and build proprietary skills
Compliance: Enterprises with strict data policies can run OpenClaw on-premises

Compare this to managed AI chat platforms that store your conversations on their servers and charge monthly fees. OpenClaw gives you the same power with full ownership.

OpenClaw in a Panaversity Context

This book is written for Panaversity students — developers learning to build and deploy AI-powered systems. OpenClaw is an ideal learning platform because:

It is real, production-grade open-source software
It covers full-stack concepts: servers, APIs, messaging protocols, AI integration
It is actively maintained and extensible
Building with it teaches patterns that apply across the AI engineering landscape

By the end of this book you will have OpenClaw installed, configured, connected to at least one messaging channel, running AI agents with custom skills, and deployed in a way your whole team can use.

Quick Reference: Key Files and Commands

Item	Value
Install command	`npm install -g openclaw@latest`
Onboard command	`openclaw onboard --install-daemon`
Config file	`~/.openclaw/openclaw.json`
Start gateway	`openclaw start`
Official docs	https://docs.openclaw.ai/
GitHub	https://github.com/openclaw/openclaw
ClawHub	https://openclawlab.com/en/docs/tools/clawhub/

Tip: You do not need to understand every concept deeply right now. The goal of this chapter is orientation — knowing the map before you walk the territory. Each concept introduced here gets its own dedicated chapter later in the book.

Next: Chapter 2 — Installation & Setup — Install OpenClaw, run the onboarding wizard, and get your gateway live in under ten minutes.