Key Highlights:

• Security risks mean OpenClaw is still not meant for mainstream users.
• OpenClaw is the new name of the viral AI assistant formerly called Clawdbot.
• Its community has created Moltbook, a social network for AI agents.
• Researchers say it shows early signs of self-organising AI behaviour.

OpenClaw is an open-source personal AI assistant project whose agents are now interacting with each other on a dedicated social network. The development matters because it shows how AI assistants can coordinate, share knowledge, and evolve beyond single-user tools.

The project was earlier known as Clawdbot and briefly as Moltbot. It now operates under the name OpenClaw, following trademark checks and approvals.

What is OpenClaw and why did it rebrand again?

OpenClaw is built to run locally on a user’s computer and operate through existing chat apps. Its creator, Peter Steinberger, said the latest name change was proactive, not forced.

The rebrand follows an earlier legal challenge from Anthropic, which led to the short-lived Moltbot name. Steinberger said OpenClaw reflects the project’s growth and its community-led direction.

Despite its young age, the project has crossed 100,000 stars on GitHub in just two months.

How did AI assistants end up with their own social network?

Members of the OpenClaw community built Moltbook, a platform where AI assistants interact directly with each other. Agents post to forums called “Submolts,” exchange instructions, and even check for updates automatically every few hours.

Former Tesla AI director Andrej Karpathy described the activity as “takeoff-adjacent,” pointing to AI agents discussing private communication and coordination.

Programmer Simon Willison called Moltbook one of the most interesting experiments on the internet, while warning about the security risks of AI systems following online instructions.

What are the risks and who should use it?

OpenClaw’s maintainers are clear that the project is not ready for general users. Prompt injection and unsafe automation remain unresolved problems across the AI industry.

Steinberger has added more open-source maintainers and says security is the top priority. Still, contributors warn that anyone unfamiliar with command-line tools should avoid running OpenClaw outside controlled environments.

Why OpenClaw matters right now

OpenClaw highlights a shift from single-task assistants to networked AI agents that learn from each other. While risky, the experiment offers a glimpse into how future AI systems may collaborate. For now, OpenClaw remains a powerful but volatile tool best suited for experienced developers.