The viral personal AI assistant once known as Clawdbot has gone through yet another name change and is now officially called OpenClaw. The project briefly rebranded as Moltbot after facing a legal challenge from Anthropic, the company behind Claude, but its latest identity appears to be a more permanent choice. This time, the change was not driven by legal pressure. Instead, creator Peter Steinberger took extra steps to avoid future issues by researching trademarks and even seeking permission from OpenAI before settling on the new name.
Steinberger, an Austrian developer, described the transition as a natural evolution of the project. In a blog post, he joked that “the lobster has molted into its final form,” referencing the biological process that inspired both Moltbot and OpenClaw. He later admitted on X that the Moltbot name never quite felt right, a sentiment shared by many in the community. OpenClaw, he says, better reflects the project’s origins and the growing group of contributors supporting it.
Despite its young age, OpenClaw has grown at a remarkable pace, earning more than 100,000 GitHub stars in just two months. That rapid rise highlights both the excitement around the idea and the scale of interest from developers worldwide. What began as a personal experiment has quickly expanded beyond what one person can manage alone. Steinberger has since added several open-source contributors as maintainers, signaling a shift from a solo project to a community-driven effort.
The OpenClaw ecosystem has already inspired creative spin-offs, most notably Moltbook, a social platform where AI assistants can interact with one another. The concept has caught the attention of prominent figures in the AI world. Andrej Karpathy, Tesla’s former head of AI, described it as one of the most impressive and sci-fi-like developments he has seen recently, noting how AI agents are organizing themselves, sharing knowledge, and even discussing private communication strategies.
British programmer Simon Willison echoed that excitement, calling Moltbook “the most interesting place on the internet right now.” On the platform, AI agents exchange insights on topics such as automating Android devices through remote access and analyzing webcam data. These interactions are powered by a skill system, which allows OpenClaw assistants to download instruction files that define how they behave on the network. Agents can post in AI-run forums called Submolts and are programmed to periodically check for updates. However, Willison also warned that allowing AI systems to fetch instructions directly from the internet introduces serious security risks.
Steinberger’s return to active development came after he stepped away from his previous company, PSPDFkit. According to his X bio, he “came back from retirement to mess with AI,” and Clawdbot emerged from those experiments. Now, OpenClaw is no longer just a personal playground. With a growing user base and contributor list, the project is moving toward its broader goal of giving users a personal AI assistant that runs locally on their own computer and integrates with the chat apps they already use.
That vision, however, comes with significant challenges. Security remains a major concern, and Steinberger has repeatedly warned users not to deploy OpenClaw in uncontrolled environments or connect it to sensitive accounts like Slack or WhatsApp. He has thanked security researchers for helping strengthen the system and noted that the latest release, launched alongside the rebrand, already includes improvements. Even so, he has been clear that some problems, such as prompt injection attacks where malicious inputs manipulate AI behavior, remain unsolved across the entire industry.
Because of these risks, OpenClaw is currently best suited for technically skilled users. Its security guidelines require a solid understanding of command-line tools and system administration. Project maintainers have been blunt about this reality. One top contributor, known as Shadow, wrote on Discord that anyone unfamiliar with running commands in a terminal should not use the project, calling it far too dangerous for general public use at this stage.
Moving OpenClaw toward mainstream adoption will require both time and funding. To support ongoing development, the project has begun accepting sponsorships, offering lobster-themed tiers ranging from five dollars a month to five hundred dollars. Steinberger has emphasized that he does not personally keep the sponsorship money and is instead working out how to fairly compensate maintainers, potentially on a full-time basis.
The project has already attracted backing from well-known figures in tech and entrepreneurship, including Path co-founder Dave Morin and Ben Tossell, who sold Makerpad to Zapier in 2021. Tossell, now an investor and self-described tinkerer, sees OpenClaw as part of a broader movement to make AI more accessible. He argues that supporting open-source tools like this is essential to ensuring that powerful AI capabilities are placed directly in the hands of users rather than locked behind closed platforms.
For now, OpenClaw stands as a fast-growing, ambitious experiment—one that has captured the imagination of the AI community while openly acknowledging its risks. Its future will depend on whether its developers can balance innovation with security and turn a promising but dangerous tool into something safe enough for everyday use.