OpenClaw is the viral AI assistant that lives on your device — what you need to know

(Image credit: OpenClaw)

When I think about what a true AI personal assistant should look like, I imagine something autonomous, capable of taking helpful actions that lead to outcomes I would have aimed for myself, and isn't a pain to communicate with.

The autonomous AI assistant OpenClaw, that’s literally only a text away, is getting scarily close to that vision.

What started out as a technologist’s weekend hobby has now garnered over 160,000 stars on GitHub. It runs locally on your machine and communicates through messaging apps you’re already using, and can pretty much pull off whatever you ask of it.

If you’re looking for something simple, polished and secure, you're 100% better off sticking with established chatbots like ChatGPT or Gemini.

If you’ve heard of OpenClaw (or maybe even its Moltbook offshoot) but weren’t sure what the fuss is about, I’ve got you covered.

What is OpenClaw?

Bear with me for a second, but I want to ensure you’re up to speed and have some background info that will come in handy later. It’s worth taking a moment to observe the evolution of OpenClaw’s name.

Austrian developer Peter Steinberger originally built the tool in late 2025 as a weekend experiment called Warelay (short for"WhatsApp Relay"), a simple bridge to let users interact with AI through their existing messaging apps. As the project grew, Steinberger rebranded it to "Clawdbot," a play on Anthropic's Claude AI model with a lobster as its mascot.

That didn't last long. Anthropic reached out with concerns about trademark confusion, and Steinberger settled on "Moltbot," a reference to lobsters shedding their shells.

But the name never quite stuck with users, and by late January 2026, the project underwent its final metamorphosis into "OpenClaw". The new name has reportedly also been okayed by OpenAI, hopefully putting an end to the confusion.

What OpenClaw actually does

Programmers and developer teams are coding and developing software — (Image credit: Getty Images)

The assistant runs entirely on your own hardware, whether that's a laptop, a home server, or a virtual private server in the cloud.

Essentially, OpenClaw represents a different approach to AI assistants like ChatGPT or Gemini.

Rather than visiting a website or downloading yet another app, users interact with OpenClaw through tools they already use daily — WhatsApp, Telegram, Discord, Slack, iMessage, and over a dozen other messaging platforms.

You don’t go to OpenClaw. You DM it.

The assistant runs entirely on your own hardware, whether that's a laptop, a home server, or a virtual private server in the cloud.

Want OpenClaw to check your calendar or reschedule a flight? It can actually do it. Send a text from your phone and OpenClaw will start opening browsers, clicking buttons, accessing files, and executing commands on your behalf. The system maintains persistent memory across sessions, remembering your preferences, past conversations, and personal context to make for an experience that’s as close to having an actual assistant as it gets.

OpenClaw isn't tied to any single AI model either. Users can plug in their own API keys for Claude, ChaGPT, Gemini, or even run local models.

There’s also a bunch of productivity integrations you can use that extend to Notion, Trello, GitHub, and email clients. Smart home enthusiasts can connect Philips Hue lights, Home Assistant hubs, and Spotify for multi-room audio control.

Why did OpenClaw blow up?

Elon Musk next to the X logo for the social media network that used to be called Twitter

(Image credit: Shutterstock)

OpenClaw went viral because its users genuinely love it and find it useful.

People have been using it to filter their emails, control their home appliances, aggregate data from wearable devices, and schedule social media posts.

It’s also fun to play around with. One X user had his OpenClaw AI bot call local businesses and uploaded the recordings.

The recordings of my OpenClaw AI bot calling local businesses are WILD (and pretty funny) pic.twitter.com/Ul7LmI5DNJFebruary 2, 2026

Secondly, as these things usually go, the hype increased when people with large followings started talking about the project.

Syracuse University professor Shelly Palmer said: “OpenClaw works exactly as advertised”. SEO expert Julian Goldie hailed it as “the new era of local AI.”

It also made it to mainstream media, with major U.S. and international publications picking up the story. So, the hype creates the hype.

Is there a link between OpenClaw and Moltbook?

What really added fuel to the fire was the birth of a new social network.

Still with me after OpenClaw’s winding naming history? Good — because there’s one more name to add to the mix: Moltbook.

In what might be the most surreal development of the OpenClaw saga, Y Combinator alum Matt Schlicht configured his OpenClaw bot, an agent he named Clawd Clawderberg, to build Moltbook, a social network designed exclusively for AI agents. The Reddit-like platform allows OpenClaw instances to create profiles, post updates, and interact with other artificial personalities.

So just like we flocked to Mark Zuckerberg's Facebook a couple of years ago, OpenClaw bots are heading to Clawderberg's Moltbook en masse (1.6 million bots at the time of writing) and talking about topics ranging from business to religion — and in some cases, even setting up their own.

On X, Elon Musk — never shy about getting involved in stuff like this — claimed we’re at “the very early stages of the singularity”.

Just the very early stages of the singularity. We are currently using much less than a billionth of the power of our Sun. https://t.co/k332z1ip7tJanuary 31, 2026

Should you try OpenClaw?

Man looking at three different computer screens at desk — (Image credit: Pexels / Olia Danilevich)

Here’s where things get slightly more complicated.

For all its promise, OpenClaw has security experts deeply worried. Cisco's threat and security research team published an analysis in late January that was pretty blunt: while acknowledging OpenClaw's capabilities as "groundbreaking" from a feature perspective, they warned that "from a security perspective, it's an absolute nightmare."

The concerns are real and significant. OpenClaw can run shell commands, read and write files, and execute scripts on your machine — capabilities that become dangerous when misconfigured or when users install community skills containing malicious code.

OpenClaw requires administrative privileges to run commands, install applications, and modify files. If your bot starts picking up hidden, rogue prompts (this is known as prompt injection), there’s not really such a thing as a break that gets triggered or a warning light that starts flashing, telling you something is off.

From a security perspective, it's an absolute nightmare.
Amy Chang, Leader, Threat & Security Research AI Software & Platform at Cisco

Researchers have already documented instances of exposed API keys, leaked email addresses, and internet-facing control panels that gave attackers full access to users' systems. Some of this data was found hidden in plain sight in plain text files, making it vulnerable to extraction.

Creator Steinberger himself admitted in a blog “this project has grown far beyond what I could maintain alone.”

If you’d still like to take the plunge and explore OpenClaw consider running it in a sandbox, far away from any of your sensitive personal or work data. That includes your main messaging accounts.

The project did, however, do a good job of capturing the imagination of the AI community and raised hopes that creating a truly autonomous and secure AI personal assistant is within reach.

It also raises interesting questions about where AI is heading. For instance, if users can run surprisingly powerful AI agents on old hardware lying in their basement (no, you don’t need to rush out and buy a Mac mini to run OpenClaw) do companies like OpenAI really need to pour billions into building what rumors suggest could be five new AI-specific devices?

Everyone's running @openclaw on Mac Minis and cloud VMs. @WesRoth just bought a new mini PC for it.I'm running it on a 13-year-old HP laptop with Ubuntu.It designed my entire Kickstarter go-to-market strategy, manages my PR pipeline, and replaced what would've been a 5-person… pic.twitter.com/acJZHXQcMAFebruary 4, 2026

OpenClaw is worth texting if you’ve got a list of repetitive, low-risk tasks and an old desktop gathering dust, just don’t trust it farther than you could throw the machine running it.

Follow Tom's Guide on Google News and add us as a preferred source to get our up-to-date news, analysis, and reviews in your feeds.

More from Tom's Guide

Christoph Schwaiger is a journalist, mainly covering technology, health, and current affairs. His stories have been published by Tom's Guide, Live Science, New Scientist, and the Global Investigative Journalism Network, among other outlets. Christoph has appeared on LBC and Times Radio. Additionally, he previously served as a National President for Junior Chamber International (JCI), a global leadership organization, and graduated cum laude from the University of Groningen in the Netherlands with an MA in journalism. You can follow him on X (Twitter) @cschwaigermt.

You must confirm your public display name before commenting

Please logout and then login again, you will then be prompted to enter your display name.