An unlikely mascot has emerged from the recent surge of interest in artificial intelligence: a lobster. The personal AI assistant known as Clawdbot captured viral attention just weeks after its debut. Although a legal challenge from Anthropic forced a name change to Moltbot, its crustacean-inspired identity remains. Before joining the growing user base, however, there are some key details to understand.

Promising to be the "AI that actually does things," Moltbot (originally Clawdbot) handles tasks like calendar management, sending messages through various apps, and even checking users in for flights. This compelling proposition has attracted thousands, despite the technical setup involved. The project began as a personal tool built by a single developer for his own needs.

That developer is Peter Steinberger, an Austrian founder known online as @steipete, who actively documents his work. After departing from his previous venture, PSPDFkit, Steinberger experienced a period of disengagement, barely using his computer for three years, as he detailed on his blog. His eventual rekindled passion led directly to Moltbot.

While Moltbot has evolved beyond a solo endeavor, its public version is rooted in "Clawd," Steinberger's personal assistant—now called Molty. He created it to help manage his digital life and explore the potential of human-AI collaboration.

The buzz surrounding Moltbot has even influenced financial markets. Cloudflare's stock jumped 14% in premarket trading on a recent Tuesday, as social media excitement about the AI agent renewed investor interest in the company's infrastructure, which developers use to run Moltbot locally.

For Steinberger, this wave of attention meant diving deeper into the AI momentum that reignited his creative drive. A self-described "Claudoholic," he initially named his project after Anthropic's flagship AI, Claude. Despite the rebranding, the project's "lobster soul" endures.

To its early users, Moltbot stands at the forefront of practical AI assistance. Enthusiasts who were already eager to use AI for building websites and apps are particularly drawn to the idea of a personal assistant that performs tasks. Like Steinberger, they enjoy experimenting with it. This enthusiasm is reflected in Moltbot rapidly accumulating over 44,200 stars on GitHub.

Nevertheless, it remains a tool for early adopters, which may be advantageous. Installation requires technical know-how, including an understanding of the inherent security risks. On one hand, Moltbot is designed with safety considerations: it is open-source, allowing code inspection, and it operates on a user's own computer or server, not in the cloud.

On the other hand, its core functionality is inherently risky. As entrepreneur and investor Rahul Sood noted, "'actually doing things' means 'can execute arbitrary commands on your computer.'"

A primary concern for Sood is "prompt injection through content," where a malicious actor could send a WhatsApp message that triggers Moltbot to perform unintended actions on a computer without the user's knowledge. This risk can be partially reduced through careful configuration. Since Moltbot supports various AI models, users can select options based on their security needs. However, complete prevention requires running the tool in an isolated silo.

This may be clear to experienced developers experimenting with a new project, but some have become more vocal in warning users drawn by the hype: careless use could lead to serious problems. Steinberger himself encountered malicious activity when he "messed up" the project's renaming process. He reported on social media that "crypto scammers" seized his GitHub username to create fake cryptocurrency projects in his name, warning followers that "any project that lists [him] as coin owner is a SCAM." He later confirmed the GitHub issue was resolved but cautioned that the only legitimate account is @moltbot, "not any of the 20 scam variations of it."

This doesn't necessarily mean curious users should avoid Moltbot entirely. However, if terms like VPS—a virtual private server, essentially a rented remote computer for running software—are unfamiliar, it may be wise to wait. This is the recommended environment for running Moltbot securely for now. Sood advised, "Not the laptop with your SSH keys, API credentials, and password manager."

Currently, safe operation involves using a separate computer with disposable accounts, which undermines the goal of having a convenient AI assistant. Resolving this security-versus-utility dilemma may require solutions outside of Steinberger's direct control.

Yet, by building a tool to address his own challenges, Steinberger has demonstrated to the developer community what AI agents can truly achieve. He has shown how autonomous AI might transition from being merely impressive to becoming genuinely useful.