How Peter Steinberger built OpenClaw

Until recently, Peter Steinberger was a developer working in relative obscurity. But in November 2025, he launched the viral AI agent platform Open Claw. Within a few months, he was receiving a hero’s welcome at a San Francisco Open Claw event where 1,300 people signed up for a 500-capacity space—a crowd that felt closer to one of Apple’s famously buzzy launches. The open-source platform allows users to create and manage personal AI agents, interacting with them through familiar messaging apps like iMessage or WhatsApp, much as they would a human assistant. But it all started organically and iteratively. As the artificial intelligence coding tools he relied on improved, Steinberger’s platform became more capable and ran more smoothly. He knew he was onto something, but getting people to understand it was another story. When Steinberger first shared his work with his 50,000 followers on X, the response was muted. “It almost felt like a challenge—why can’t I explain how awesome this is?” Steinberger says. When he showed friends what his agent could do—continuously work on complex tasks—they immediately wanted their own. Realizing that showing worked better than telling, Steinberger set up a public Discord group chat as a showcase for OpenClaw. “I just worked in public, and people could observe me improving my agent . . . and doing random things like automating my house and just showing people how good this thing is, even if you throw weird problems at it.” That’s when OpenClaw began to take off. Many early fans were AI researchers, but plenty were not. (One early user was a dentist.) “I thought a little bit about marketing, but I never imagined that it would go off like this,” he says. For users willing to give an AI agent access to files on their computer, OpenClaw offered an early taste of a humanlike assistant. People built small agent-run businesses; one user had an agent contact local dealerships to find the best price on a new truck. Others created research assist