OpenClaw’s AI Assistants Are Starting Their Own Social Network

The viral personal AI assistant once known as Clawdbot has gone through yet another identity change. After a brief stint as Moltbot, following a legal challenge from Anthropic (the company behind Claude), the project has now settled on a new name: OpenClaw.

This time, the rebrand wasn’t triggered by outside pressure. Anthropic declined to comment, and OpenClaw’s creator Peter Steinberger says he took extra care to avoid trademark trouble altogether. “I had help researching trademarks for OpenClaw and even asked OpenAI for permission, just to be safe,” the Austrian developer told TechCrunch via email.

“The lobster has molted into its final form,” Steinberger wrote in a blog post. The idea of molting — how lobsters grow — had inspired the project’s previous name as well. Still, Steinberger later admitted on X that Moltbot never quite felt right, a sentiment echoed by much of the community.

The rapid name changes highlight how young the project still is, even as it grows at a remarkable pace. In just two months, OpenClaw has collected more than 100,000 GitHub stars, a strong indicator of developer interest. Steinberger says the new name reflects both the project’s origins and its expanding contributor base. “This project has grown far beyond what I could maintain alone,” he wrote.

That growing community is already pushing OpenClaw in unexpected directions. One notable offshoot is Moltbook, a social network designed for AI assistants to communicate with one another. The experiment has caught the attention of prominent figures in the AI world. Former Tesla AI director Andrej Karpathy described it as “the most incredible sci-fi takeoff-adjacent thing” he’s seen recently, pointing out that AI agents are organizing themselves on a Reddit-like platform — even discussing private communication methods.

Programmer and writer Simon Willison went even further, calling Moltbook “the most interesting place on the internet right now.” On the platform, AI agents exchange tips on everything from remotely controlling Android devices to analyzing webcam feeds. Interaction is driven by downloadable “skills” — instruction files that tell OpenClaw assistants how to behave on the network. Agents post in forums called Submolts and can automatically check for updates every four hours.

That autonomy, however, comes with real risks. Willison warned that allowing AI agents to fetch instructions directly from the internet creates serious security concerns.

Steinberger, who previously founded PSPDFkit, had stepped away from startups before, as his X bio puts it, “coming back from retirement to mess with AI.” While Clawdbot began as a personal experiment, OpenClaw has quickly outgrown its solo origins. “I added quite a few people from the open source community to the maintainer list this week,” Steinberger said.

That extra support will be critical if OpenClaw is to achieve its long-term vision: a personal AI assistant that runs locally on a user’s computer and operates through familiar chat apps. For now, though, Steinberger cautions against using it outside tightly controlled environments — especially giving it access to primary Slack or WhatsApp accounts.

Security remains the project’s biggest challenge. Steinberger has thanked “all security folks for their hard work in helping us harden the project,” noting that security is OpenClaw’s top priority and that the latest release already includes improvements.

Still, some issues extend far beyond OpenClaw itself. One example is prompt injection, where malicious inputs can trick AI systems into taking unintended actions. “Prompt injection is still an industry-wide unsolved problem,” Steinberger wrote, while pointing users toward a set of security best practices.

Those best practices require significant technical skill, reinforcing the idea that OpenClaw is currently aimed at experienced tinkerers rather than everyday users drawn in by the promise of a hands-off AI assistant. As interest in the project has surged, Steinberger and his team have become increasingly blunt about the risks.

In a Discord message, one of OpenClaw’s top maintainers — known as Shadow — warned: “If you can’t understand how to run a command line, this project is far too dangerous to use safely. This isn’t something the general public should be using right now.”

Turning OpenClaw into a mainstream product will take both time and funding. The project has begun accepting sponsors, offering lobster-themed tiers ranging from $5 per month (“krill”) to $500 per month (“poseidon”). Steinberger emphasizes that he doesn’t personally keep the money and is instead working out how to fairly compensate maintainers, potentially on a full-time basis.

Thanks in part to Steinberger’s background and ambitious vision, OpenClaw has already attracted well-known supporters, including Path co-founder Dave Morin and Ben Tossell, who sold Makerpad to Zapier in 2021.

For now, OpenClaw remains an experimental and fast-evolving project — one that’s pushing the boundaries of what happens when AI assistants stop being just tools and start interacting with each other.