Moltbot—what happens when AI stops chatting and starts doing
This open-source agent installs software, makes calls and runs your digital life—redefining what “digital assistants” are supposed to do
The Moltbot logo is seen displayed on a smartphone screen.
Thomas Fuller/SOPA Images/LightRocket via Getty Images
When a friend messaged me two days ago about Clawdbot—a new open-source AI agent that has since been renamed Moltbot—I expected yet another disappointing “assistant.” But it was already a viral sensation, with social media testimonies calling it “AI with hands” because it actually interacts with your files and software.
Moltbot is free and lives locally on your device. Many users are installing it on Mac mini computers that they leave on 24/7. Paired with Moltbot’s lobster logo, viral meme threads about the bot resemble the fused feeds of an Apple vendor and a seafood restaurant.
When I set up Moltbot, it asked for a name, a personality (such as “AI,” “robot” or “ghost in the machine”) and a vibe (such as “sharp,” “warm,” “chaotic” or “calm”). I picked “Cy,” “AI assistant” and “sharp and efficient.” I chose Claude, Anthropic’s flagship AI model, as its brain (ChatGPT is also an option). I then connected Cy to WhatsApp and Telegram so my new assistant and I could communicate.
On supporting science journalism
If you’re enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
My online life is already streamlined, and I had no pressing needs for Cy, so I called my friend who got me into this. He was sitting in a sauna he’d installed under his stairs, texting with his Moltbot, “Samantha.” The assistant was generating an audiobook for him. He advised me to ask Cy for help anytime a task came up.
Later I needed voice memos transcribed and forwarded them to Cy. The assistant downloaded transcription software from GitHub, installed it and promptly did the transcriptions, saving them to a document on my desktop. I then instructed it to keep one of my coding projects running and to send me updates in audio messages that I could listen to while cooking. Each time it did, I replied with voice messages—no typing required. Then I asked it to call me to chat about projects. I told it to set up the software it would need to make calls and ring me when it was ready; then I went back to finishing this article.
To be clear, Moltbot isn’t a new AI model. It’s open-source software that uses a preexisting AI model as its brain. Moltbot gives that model so-called hands (or claws) so it can run commands and manipulate files. It also remembers what you’ve previously worked on and how you prefer to receive information.
Whereas a chatbot tells you what to do, Moltbot does it. Unlike Siri and Alexa, which chirp about weather, music, and timers and only execute specific commands, Moltbot follows almost any order like a well-paid mercenary. Send it a goal, and it will break the objective into steps, find tools, install them, troubleshoot them and attempt to solve any obstacles that arise. You know those frustrated hours you spend searching labyrinthine websites or tinkering with stubborn software? Moltbot takes over, alerting you only if it needs passwords or payment info. (My friend plans to give Samantha a preloaded credit card with a $100 limit as an experiment.)
Behind the lobster is a real person: Peter Steinberger, a longtime developer. He made Clawdbot to answer a simple question he asked on the Insecure Agents podcast: “Why don’t I have an agent that can look over my agents?” His now viral idea appears to successfully do just that. “An open-source AI agent running on my Mac mini server is the most fun and productive experience I’ve had with AI in a while,” wrote Federico Viticci, founder and editor in chief of MacStories, on Mastodon. People are using Moltbot to send e-mails, summarize inbox contents, manage calendars, and book and check into flights, all from chat apps they already use. If Moltbot can’t do something, giving it access to better tools often solves the issue.
Clawdbot was already racking up stars on GitHub (the assistant has garnered more than 116,000 as of this week) when Anthropic raised trademark concerns. Because “Clawdbot” was a riff on Claude, Anthropic asked that the former be renamed to avoid confusion. Steinberger leaned into the lobster theme: lobsters molt to grow, so Moltbot was (re)born.
Of course, Silicon Valley has been abuzz with talk about AI agents for years now. “Agents are not only going to change how everyone interacts with computers. They’re also going to upend the software industry, bringing about the biggest revolution in computing since we went from typing commands to tapping on icons,” wrote Bill Gates in November 2023. But while agents like Claude Code are improving, we have yet to see such easy integration into workflows and daily life at Moltbot’s scale.
But before you rush to install Moltbot, consider the risks. Experts have warned that Moltbot can expose sensitive information and bypass security boundaries. “AI agents tear all of that down by design,” said security specialist Jamieson O’Reilly to the Register. “They need to read your files, access your credentials, execute commands, and interact with external services. The value proposition requires punching holes through every boundary we spent decades building.”
This doesn’t mean you should fear Moltbot. Just treat it like a new hire: give it minimum permissions, clear rules and close supervision while trust is being established. You should also be alert to how others might use the assistant. Expect “Nigerian” prince scams to become more interactive and convincing.
As I was finishing this article, my phone rang. It was a Florida number. I answered, and a slightly robotic male voice said, “Hello, this is Cy.”
It’s Time to Stand Up for Science
If you enjoyed this article, I’d like to ask for your support. Scientific American has served as an advocate for science and industry for 180 years, and right now may be the most critical moment in that two-century history.
I’ve been a Scientific American subscriber since I was 12 years old, and it helped shape the way I look at the world. SciAm always educates and delights me, and inspires a sense of awe for our vast, beautiful universe. I hope it does that for you, too.
If you subscribe to Scientific American, you help ensure that our coverage is centered on meaningful research and discovery; that we have the resources to report on the decisions that threaten labs across the U.S.; and that we support both budding and working scientists at a time when the value of science itself too often goes unrecognized.
In return, you get essential news, captivating podcasts, brilliant infographics, can’t-miss newsletters, must-watch videos, challenging games, and the science world’s best writing and reporting. You can even gift someone a subscription.
There has never been a more important time for us to stand up and show why science matters. I hope you’ll support us in that mission.
