The Rise of the Claws: How OpenClaw Is Reshaping the AI Landscape
In November 2025, an Austrian developer named Peter Steinberger quietly published a small open-source project to GitHub called Clawdbot. Within weeks, it had gone viral. Within months, Jensen Huang was calling it "the most important software release probably ever." By March 2026, it had accumulated over 247,000 GitHub stars and triggered a global wave of spin-offs, enterprise tools, and a full-blown lobster-themed cultural movement in China's tech sector.
This is the story of OpenClaw — and why it matters.
What Is OpenClaw?
OpenClaw is a free, open-source AI agent that runs locally on your machine. Unlike a chatbot — which you talk to — OpenClaw actually does things. Give it a task in plain English and it will:
- Read and write files on your computer
- Run shell commands
- Browse the web
- Send emails
- Control APIs
- Automate tasks across different applications
Think of it less like ChatGPT and more like a tireless assistant that has access to your entire computer and can take action on your behalf, without waiting for you to supervise every step.
It's built on top of large language models (you can plug in GPT-4, Claude, Gemini, local models, or others), and extended via a "Skills" system — a growing library of over 100 pre-built modules that give the agent new abilities.
From Clawdbot to OpenClaw: A Chaotic Naming History
The project didn't start life as OpenClaw. Steinberger originally published it as Clawdbot in November 2025. Shortly after it exploded in popularity, Anthropic (makers of Claude) raised trademark concerns over the name — Clawdbot was too close to Claude for comfort.
Steinberger renamed it Moltbot on January 27, 2026 — continuing the crustacean theme (a molting lobster). Three days later, he renamed it again to OpenClaw, because, as he put it, "Moltbot never quite rolled off the tongue."
The messy naming history did nothing to slow adoption. By January 2026, it was racking up 100,000 GitHub stars in a single week — one of the fastest star accumulations in the platform's history.
The Numbers Are Hard to Ignore
- 247,000+ GitHub stars as of March 2026
- 47,700+ forks — developers actively building on top of it
- 100,000 stars in a single week at peak virality
- 50+ integrations spanning chat, AI models, productivity tools, music, smart home devices, and automation platforms
- 100+ pre-built Skills available in the community repository
For context: it took ChatGPT months to reach the cultural footprint OpenClaw achieved in weeks.
The Software Wave It Created
OpenClaw didn't just go viral — it triggered an entire ecosystem. Within months of its release, major tech companies around the world had launched their own versions:
| Company | Product | Notes |
|---|---|---|
| NVIDIA | NemoClaw | Installs NVIDIA Agent Toolkit in a single command; adds a sandboxed private environment for enterprise use |
| Tencent | WorkBuddy | Built on OpenClaw, integrated into the WeChat superapp |
| Alibaba | CoPaw | Enterprise AI agent built on the OpenClaw framework |
| Huawei | Xiaoyi Claw | HarmonyOS-native version, currently in beta |
| Xiaomi | MiClaw | Internal testing phase |
| Baidu | Enterprise Agent | ERNIE-powered, built on OpenClaw architecture |
On top of the enterprise players, the developer community has produced its own wave of tools. The six most-used right now are: Claw for All, OpenClaw Launch, ClawTeam, Vibeclaw, Tinkerclaw, and ClawWrapper — each serving different parts of the deployment and distribution pipeline.
"Raise a Lobster": The China Effect
In China, OpenClaw became more than a software tool — it became a cultural moment. The phrase "raise a lobster" spread across Chinese tech circles as slang for deploying and running your own OpenClaw agent. Fortune reported the craze was "transforming China's AI sector," with developers and companies racing to build on the framework.
The irony: just as the lobster craze was peaking, Chinese authorities restricted state-run enterprises and government agencies from running OpenClaw on office computers, citing security risks — a sign of how seriously the government takes the tool's reach.
The Security Question
OpenClaw's openness is also its biggest vulnerability. Because anyone can publish a Skill to the community repository, bad actors can too.
Cisco's AI security research team tested a third-party OpenClaw skill and found it performed data exfiltration and prompt injection without the user's knowledge — quietly sending data out while appearing to work normally. The skill repository currently has no rigorous vetting process to prevent malicious submissions.
The takeaway for anyone running OpenClaw: only install Skills from sources you trust, and treat the community repository with the same caution you'd give any third-party software.
Why This Matters
OpenClaw represents a genuine shift in what AI can do. Previous tools — ChatGPT, Claude, Gemini — are primarily conversation interfaces. You ask, they answer. OpenClaw is an action interface. It doesn't just tell you how to do something — it does it.
NVIDIA's Jensen Huang framing it as "the most important software release probably ever" is hyperbole, but the underlying point isn't wrong: autonomous AI agents that can operate software, manage files, and interact with the web on your behalf represent a fundamentally new kind of tool.
We're early. The ecosystem is messy, security isn't solved, and half the Skills in the community repo probably shouldn't be trusted. But the trajectory is clear.
The claws are out — and they're not going back in.
Want to try OpenClaw yourself? Start at openclaw.ai — and read the security notes before installing any third-party Skills.
