What is OpenClaw (Clawdbot)? The AI Assistant the Internet Cannot Stop Talking About

The Project That Broke the Internet

In January 2026, a developer named Peter Steinberger published an open-source project called Clawdbot. Within weeks it had 29,000 GitHub stars. Within a month it was trending globally, causing Mac Mini shortages, and being discussed by tech investors and startup founders on every platform.

By February 15, 2026 — just weeks after launch — OpenAI CEO Sam Altman announced that Steinberger was joining OpenAI. The project, by then renamed OpenClaw, had 145,000 GitHub stars and 20,000 forks.

Here is what it actually is and why it captured the world's imagination.

What is OpenClaw?

OpenClaw (originally Clawdbot, briefly Moltbot) is an open-source, self-hosted personal AI assistant. The simplest description: it is Claude — or GPT-4, or DeepSeek, or Gemini — with hands.

Standard AI assistants like ChatGPT live in a browser tab. You type, they respond, the conversation ends. OpenClaw is different in three fundamental ways:

It lives in your messaging apps. OpenClaw integrates with WhatsApp, Telegram, Discord, Slack, Signal, and iMessage. Your AI assistant is in the same app where you talk to your friends and colleagues. Same conversation, same memory, everywhere.

It runs 24/7. A standard AI session ends when you close the tab. OpenClaw runs on your own hardware — a Mac Mini, a cheap VPS, an old laptop — continuously. It is always on, always available.

It actually does things. Most AI assistants generate text. OpenClaw executes actions: browsing the web, sending emails, managing calendar events, executing terminal commands, controlling smart home devices, checking you in for flights. It acts on your behalf, not just advises.

Who Built It and Why

Peter Steinberger is an Austrian developer, founder of PSPDFKit (now Nutrient), a company that builds PDF processing tools used by thousands of enterprise companies. He came out of semi-retirement to build OpenClaw after documenting his personal AI workflow publicly — including a widely shared blog post titled "Claude Code is my computer."

His philosophy: AI assistants should be ambient, persistent, and capable of acting — not just conversational. He built what he wanted to use himself, open-sourced it, and watched it go viral.

What OpenClaw Can Actually Do

Users have documented OpenClaw performing tasks including:

• Email management — reading, summarising, drafting, and sending emails autonomously
• Calendar management — scheduling meetings, setting reminders, managing conflicts
• Web browsing — researching topics, summarising articles, monitoring websites for changes
• Flight check-in — automatically checking in for flights when the window opens
• Smart home control — adjusting lights, temperature, and connected devices
• Terminal command execution — running scripts, managing files, automating development tasks
• Morning briefings — proactively sending you a summary of news, weather, and your schedule
• PDF summarisation — reading and summarising documents you share with it

The key distinction from tools like Zapier or IFTTT is intelligence. OpenClaw does not follow rigid rules — it understands context, reasons about what you need, and executes accordingly.

The Drama: Clawdbot to Moltbot to OpenClaw

The project's name changed three times in rapid succession, and each change came with a story.

Clawdbot — the original name. A play on "Claude" (the AI) and "bot." This is what went viral.

The C&D — Anthropic, which makes Claude, issued a cease-and-desist letter over the name Clawdbot, arguing it was too close to their Claude brand. Steinberger complied and renamed the project.

Moltbot — the brief rename. Within days, crypto scammers launched a Moltbot token and flooded search results with scam content. Legitimate users could not find the real project. The chaos became a news story in itself.

OpenClaw — the final name. Steinberger settled on OpenClaw, moved the project to a new GitHub organisation, and it continued growing. The accumulated drama — C&D, crypto scammers, viral coverage — paradoxically made the project more famous than it would have been otherwise.

Peter Steinberger Joins OpenAI

On February 14, 2026, Steinberger announced he was joining OpenAI. Sam Altman confirmed it publicly, saying Steinberger would work on "smart agents" — OpenAI's term for AI systems that can take autonomous actions in the world.

OpenClaw itself was moved to an open-source foundation. Altman stated that OpenAI would continue to support it as an open-source project. The project will not disappear — it will be maintained by the community and backed by OpenAI's support.

The hire made sense from OpenAI's perspective. Steinberger had built, in weeks, the most popular open-source AI agent in history. He had demonstrated that there was massive demand for persistent, action-capable AI assistants — exactly the direction OpenAI's product roadmap was heading.

How OpenClaw Works Technically

OpenClaw is not itself an AI model. It is an orchestration layer — software that connects your messaging apps to AI models and gives those models the ability to take actions.

The architecture:

• You send a message on WhatsApp (or Telegram, Discord, etc.)
• OpenClaw receives it via a messaging bridge
• It passes your message to an AI model — Claude, GPT-4, DeepSeek, or Gemini, depending on your configuration
• The AI model generates a response and, if action is needed, instructs OpenClaw which tools to use
• OpenClaw executes those tools (browser, email, calendar, terminal)
• The result comes back to you in WhatsApp

The AI model is the brain. OpenClaw is the hands and the persistent memory.

Why It Went Viral

Three things combined:

The "Jarvis" fantasy. Since Iron Man, the cultural dream of a personal AI that actually does things — not just talks — has been waiting for a product to fulfil it. OpenClaw was the first open-source project that came close enough to feel real.

Timing. It launched at a moment when AI capability had jumped dramatically. Claude, GPT-4, and DeepSeek were all capable enough to actually execute complex tasks reliably. The same project would not have worked as well in 2023.

Open source + self-hosted. People are increasingly uncomfortable with their AI assistant being a closed product run by a US company with access to their messages, emails, and calendar. OpenClaw's self-hosted model — your data stays on your hardware — resonated strongly with privacy-conscious developers globally.

Should You Set It Up?

OpenClaw is genuinely powerful. It is also genuinely complex. Setting it up requires a server (VPS or local hardware), comfortable with command line, and willingness to manage your own infrastructure.

If you are a developer comfortable with Linux and a terminal, the setup is manageable. If you are not technical, the hosted version — OpenClaw Cloud — is an easier entry point, though it sacrifices the self-hosted privacy benefit.

For a step-by-step setup guide, read the OpenClaw VPS setup guide.

The more interesting question is not whether you should set it up — it is what it represents. OpenClaw is the first widely used demonstration that people genuinely want persistent, action-capable AI assistants. That demand exists. The question now is what form it takes as the major AI companies build their own versions.

OpenAI hiring its creator is a fairly clear answer to that question.