What is OpenClaw?
OpenClaw is a free, open-source autonomous AI agent that runs locally on your machine and communicates with you through messaging platforms you already use—WhatsApp, Discord, Slack, Telegram, iMessage, and more. Instead of switching to a chat window every time you need AI help, OpenClaw brings an AI agent directly into your daily conversations.
In this lesson you will learn where OpenClaw came from, what makes it different from a chatbot, and which AI models and platforms it supports.
A Brief History: Three Names in Three Days
OpenClaw holds the unofficial record for the fastest triple rebrand in open-source history.
| Name | Dates | What Happened |
|---|---|---|
| Clawdbot | Nov 2025 – Jan 27, 2026 | Peter Steinberger (founder of PSPDFKit) built the first version in a single night, connecting a chat app with Claude. The name referenced the lobster mascot visible while reloading Claude Code. |
| Moltbot | Jan 27–29, 2026 | Anthropic issued trademark complaints. Steinberger cooperated and renamed to "Moltbot" (Molting Robot)—because lobsters need to shed their shells to grow. |
| OpenClaw | Jan 30, 2026 – present | Steinberger felt "Moltbot never quite rolled off the tongue" and renamed again. During the transition, handle snipers grabbed social media accounts within seconds, and crypto scammers launched a fake token that briefly hit a $16 million market cap. |
Within 72 hours of its initial launch, the project collected over 60,000 GitHub stars. It now has over 150,000 stars and 20,000+ forks, making it one of the fastest-growing open-source projects in GitHub history.
Chatbot vs. Autonomous Agent
A standard AI chatbot waits for your message, generates a reply, and stops. OpenClaw does more:
| Capability | Chatbot | OpenClaw Agent |
|---|---|---|
| Responds to messages | Yes | Yes |
| Runs as a background process | No | Yes (Gateway daemon) |
| Executes tools and APIs | No | Yes (via skills) |
| Remembers across sessions | Limited | Yes (MEMORY.md, JSONL transcripts) |
| Proactively checks tasks | No | Yes (heartbeat system) |
| Works across platforms | One interface | WhatsApp, Discord, Slack, Telegram, and more |
The key difference is autonomy. OpenClaw can plan multi-step actions, execute them using skills, and loop until the task is complete—all without you pasting instructions into a chat window.
Key Concepts at a Glance
Before diving deeper in later lessons, here are the core concepts you will encounter throughout this course:
- Gateway: A single long-lived Node.js process that coordinates everything—channel connections, session state, the agent loop, model calls, tool execution, and memory persistence.
- Channels: Adapters that connect the Gateway to messaging platforms (WhatsApp, Discord, Slack, etc.).
- Skills: Modular plugins defined in
SKILL.mdfiles that teach the agent how to perform specific tasks—from running shell commands to controlling smart home devices. - Memory: Plain Markdown and YAML files (
MEMORY.md, JSONL transcripts) that give the agent persistent context across conversations. - Heartbeat: A periodic check (every 30 minutes by default) where the agent reads a checklist and decides whether any item needs action.
- Agentic Loop: The cycle of input, planning, tool execution, and response that lets the agent handle complex multi-step tasks.
Supported AI Models
OpenClaw is model-agnostic. You bring your own API key and choose any supported provider:
| Provider | Example Models | Env Variable |
|---|---|---|
| Anthropic | Claude Opus 4.6, Claude Sonnet 4.5 | ANTHROPIC_API_KEY |
| OpenAI | GPT-5.1, GPT-4o, o1 | OPENAI_API_KEY |
| Gemini 3 Pro, Gemini 2.5 Flash | GEMINI_API_KEY | |
| DeepSeek | V3, R1 | DEEPSEEK_API_KEY |
| xAI | Grok 3, Grok 4 | XAI_API_KEY |
| OpenRouter | Access multiple providers | OPENROUTER_API_KEY |
| Local (Ollama) | Llama, Mistral, etc. | Auto-detected at localhost:11434 |
Models are referenced using a provider/model-id format (e.g., anthropic/claude-opus-4-6). You can switch models at any time with openclaw models set.
The documentation recommends using Anthropic's Opus models because they are strongest at recognizing prompt injections. Smaller or cheaper models are more susceptible to tool misuse.
Supported Platforms
OpenClaw connects to your messaging platforms through built-in channel adapters:
Built-in channels:
- WhatsApp (most popular, QR pairing via Baileys)
- Telegram (Bot API via grammY)
- Discord (Bot API + Gateway)
- Slack (Bolt SDK)
- Signal (signal-cli)
- iMessage (via BlueBubbles on macOS)
- Microsoft Teams (Bot Framework)
- WebChat (built-in web interface)
Extension channels are available for Matrix, Google Chat, Mattermost, IRC, Feishu/Lark, and more.
OpenClaw also supports voice interaction on macOS, iOS, and Android, plus a visual "Live Canvas" workspace.
Why OpenClaw Matters
OpenClaw represents a shift from "AI you visit" to "AI that works for you." It runs on your hardware, stores data in plain text files you control, and integrates with the tools you already use. Whether you want an assistant that checks your calendar every morning, monitors a GitHub repository, or drafts responses in Slack, OpenClaw provides the infrastructure to make it happen.
Key Takeaway
OpenClaw is an open-source autonomous AI agent that runs locally, connects to your messaging platforms, and executes multi-step tasks using skills. It supports every major AI model provider and stores everything in human-readable files. In the next lesson, you will install OpenClaw and run the onboarding wizard.