What Is OpenClaw?

OpenClaw is an open-source, self-hosted personal AI agent. It runs as a background process on your own machine, connects to messaging apps you already use, and responds to messages using the AI model of your choice — Claude, GPT-4, or Gemini. Your conversations and personal context stay on your hardware. There is no subscription, no cloud dependency, and no third party processing your data.

What OpenClaw does

When you send a message to your OpenClaw bot on Telegram or WhatsApp, the following happens:

1. The Gateway receives the message from the channel
2. The Agent loads your personal context (SOUL.md, USER.md, MEMORY.md)
3. The message and your context are sent to your chosen AI model's API
4. The response streams back to the messaging app

This all happens on your machine in a few seconds. The only external call is to the AI provider's API — everything else is local.

OpenClaw can also run proactively: a cron automation sends you a morning briefing at 7am, a reminder before a meeting, or a nightly summary — without you initiating anything.

The five things that define OpenClaw

1. It lives in your messaging apps. You interact with OpenClaw through Telegram, WhatsApp, Slack, Discord, or iMessage — the same apps you use to message people. No browser tab, no separate app to open.

2. Your data stays on your hardware. Conversations, context files, and memory all live on your machine. The only data that leaves is the message payload sent to your AI provider's API — which you choose and pay for directly.

3. It knows who you are. SOUL.md defines the agent's personality. USER.md stores facts about you. MEMORY.md grows across sessions. The agent reads these on every conversation, giving it persistent, editable context.

4. It can message you first. Cron automations let OpenClaw send you scheduled messages: a morning briefing, a weekly summary, a reminder. Hosted AI services like ChatGPT only respond when you initiate.

5. You choose the AI model. OpenClaw works with Claude, GPT-4, and Gemini. You can switch models by changing two lines in the config, or run different agents on different models simultaneously.

How OpenClaw compares to alternatives

Full comparisons: OpenClaw vs ChatGPT · OpenClaw vs n8n · OpenClaw vs Zapier · see all

Who OpenClaw is for

OpenClaw is for people who:

• Want an AI assistant they can message from their phone without switching apps
• Care about data privacy and don't want conversations processed by a third-party cloud service
• Want persistent memory — an agent that actually remembers what you've talked about
• Want to pay for AI by usage (API cost) rather than a fixed monthly subscription
• Are comfortable with a 15-minute command-line setup

It's not for people who need zero setup (use ChatGPT), want to automate workflows between 50 SaaS apps (use n8n or Zapier), or are building AI products for other users (use the OpenAI Assistants API).

How to get started

The quickest path:

1. Install: npm install -g openclaw (requires Node.js 18+)
2. Initialize: openclaw init — sets up your config with your AI provider key
3. Connect a channel: Add Telegram (2 minutes) or another channel
4. Start the gateway: openclaw gateway start
5. Test: Send your bot a message

The full course at howopenclaw.com/course walks through every step with no assumed knowledge — from install to automations in about an hour.

OpenClaw terminology

A few terms you'll see throughout the documentation:

• Gateway — the always-running background process that manages channels, the agent, and cron jobs
• Agent — the conversational AI component that reads your context and generates responses
• SOUL.md — a Markdown file defining your agent's personality and communication style
• USER.md — facts about you that the agent always loads
• MEMORY.md — persistent notes that grow across sessions
• Skills — extensions that add capabilities (web search, email, calendar, GitHub)
• Channels — the messaging platforms the agent communicates through

Full definitions are in the Key Concepts reference.

---