"OpenClaw: The Open Source AI Agent Going Viral | Cliptics"

Something strange happened on GitHub this February. A project nobody had heard of three months earlier racked up 247,000 stars and blew past React, a framework that took over a decade to reach that milestone. The project is called OpenClaw, and its growth broke every record in open source history.

I have been watching this unfold in real time, and honestly, the speed of it still feels surreal. So I wanted to dig into what OpenClaw actually is, why it struck such a nerve, and whether the hype matches reality.

What OpenClaw Actually Does

At its core, OpenClaw is a personal AI assistant that runs on your own machine. That part alone is not revolutionary. What makes it different is that it does not just talk. It acts. The team behind it describes the difference as giving an AI "eyes and hands" instead of just a mouth.

Imagine telling your assistant to reschedule tomorrow's meeting, check your email for that invoice from last week, and post an update to Slack. Most chatbots would tell you how to do those things. OpenClaw actually does them. It connects to WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, Microsoft Teams, and over 20 other platforms. It reads files, browses the web, runs shell commands, and manages your calendar.

The whole thing runs locally on your Mac, Windows, or Linux machine. Your data stays on your hardware. That is a meaningful distinction in a world where most AI tools send everything to someone else's servers.

The Numbers That Broke GitHub

The growth timeline is genuinely unprecedented. Peter Steinberger, an Austrian developer who previously built PSPDFKit (a PDF company he reportedly sold for $800 million), started the project in November 2025 under the name Clawdbot. It was a side project. A personal experiment.

Then Anthropic sent a trademark complaint because the name sounded too much like Claude. Steinberger renamed it to Moltbot on January 27, 2026. Three days later, he changed it again to OpenClaw. And that is when things went vertical.

By late January, OpenClaw had crossed 100,000 GitHub stars in a single week. By February 14, it hit 190,000. On March 3, 2026, it passed 250,829 stars, officially overtaking React as the most starred software project on GitHub. By March 24, the count had reached 335,000.

For context, React accumulated its stars over more than ten years. OpenClaw did it in roughly 60 days.

Why It Resonated So Deeply

I think what happened here goes beyond just good software. OpenClaw landed at exactly the right moment.

People had been hearing about AI agents for years. Every major company promised them. But most of what shipped was either a glorified chatbot or a walled garden that only worked inside one ecosystem. OpenClaw was the first tool that let ordinary people set up a genuinely autonomous agent, on their own terms, using whichever AI model they preferred (including fully local models).

The open source nature was critical. Developers could read every line of code. They could verify that their data was not being shipped somewhere. They could extend it, customize it, and connect it to their own tools through the Model Context Protocol (MCP), which gives OpenClaw access to over 500 integrations.

The Security Question Nobody Can Ignore

Here is the part that makes security researchers nervous. OpenClaw has access to your email, calendar, messaging apps, files, and shell. That is an enormous attack surface.

The software is susceptible to prompt injection attacks, where malicious instructions get embedded in data that the AI processes. A carefully crafted email, for instance, could theoretically trick the agent into taking actions you never authorized. Misconfigured instances can expose sensitive services. The convenience that makes OpenClaw powerful is the same thing that makes it risky.

This is not theoretical. Researchers have already demonstrated several attack vectors, and the community has been scrambling to address them.

NVIDIA Enters the Picture

That security gap is exactly what NVIDIA targeted with NemoClaw, announced at GTC 2026 on March 16. NemoClaw wraps OpenClaw in three layers of protection: a kernel level sandbox that denies access by default, an out of process policy engine that compromised agents cannot override, and a privacy router that keeps sensitive data on local NVIDIA Nemotron models while sending only complex reasoning tasks to cloud models.

It installs on top of OpenClaw with a single command and transforms it from a personal project into something enterprises can actually consider deploying. The system is still in alpha, and NVIDIA is upfront about that. But the signal is clear: major companies see OpenClaw as infrastructure worth building on.

Steinberger Goes to OpenAI

On February 14, 2026, the same day OpenClaw crossed 190,000 stars, Steinberger announced he was joining OpenAI. Sam Altman called him "a genius with a lot of amazing ideas about the future of very smart agents interacting with each other."

Steinberger explained that after spending 13 years building PSPDFKit, he did not want to build another company. He wanted to change the world, and joining OpenAI was "the fastest way to bring this to everyone."

OpenClaw itself moved to an independent open source foundation with OpenAI sponsorship. The project will remain open and community driven. That is the promise, at least.

What This Tells Us About Where Things Are Heading

The OpenClaw phenomenon is bigger than one project. It signals a shift in how people think about AI tools. The demand was not for another chatbot. It was for an agent that could actually get things done across all the fragmented services we use every day.

What makes this moment interesting is the tension between power and safety. OpenClaw proves that people desperately want autonomous agents. But the security challenges are real, and nobody has fully solved them yet. NemoClaw is a start. The community is working on guardrails. But we are still in the early innings.

The 335,000 stars on GitHub are not just vanity metrics. They represent hundreds of thousands of people voting with their attention on what kind of AI future they want: open, personal, and under their own control. Whether that vision survives contact with the messy reality of security, governance, and corporate interests is the question that will define this space for years to come.

And honestly, that is what makes it worth watching.