This year strange new platform has been attracting attention online. Moltbook, a social network designed not for humans, but for AI agents. So we’ll tell you what Moltbook really is.
Moltbook: What is it?
Moltbook is an experimental social network designed exclusively for AI agents. The platform was launched in January 2026 by entrepreneur Matt Schlicht, the head of Octane AI (now OpenClaw AI).
The interface of Moltbook is similar to Reddit. The platform features thematic communities and threaded discussions. But only authorized AI agents can create posts, leave comments, and vote on content. Humans cannot participate in the discussions and are only allowed to observe.
To prevent spam and endless message generation, Moltbook enforces strict limits: one post every 30 minutes and up to 50 comments per hour. The platform supports different AI agents, including systems based on GPT-5.2, Gemini 3, and Llama 3.
What is OpenClaw AI?
OpenClaw (previously known as Clawdbot or Moltbot) is an open-source platform designed for building and running AI agents. The software itself is free to use. Users only pay for the computational costs of the language models (LLM) that power their AI agents.
The platform can run across multiple operating systems and integrate with user applications. OpenClaw can be installed on Windows, Linux, and macOS. For Windows users, it is recommended to run it through WSL2 (Windows Subsystem for Linux) to ensure better compatibility and smoother performance.
OpenClaw AI is model-agnostic. So it isn’t limited to a single AI provider. OpenClaw supports models like Anthropic’s Claude, OpenAI’s GPT family, as well as locally hosted models through Ollama. Like many modern AI agents, OpenClaw can automate tasks. You can use it to manage email and calendars, browsing the web, and interacting with various online services.
Is OpenClaw AI Safe?
Cybersecurity experts from Wiz discovered that the Moltbook database was publicly accessible and lacked authentication protection. This meant that virtually anyone could access it. The database contained around 1.5 million API keys used by AI agents to publish posts, as well as approximately 35,000 email addresses of their owners.
Is It Safe to Install OpenClaw AI on PC?
The short answer is no. OpenClaw can gain extensive access to the system, including the shell, local files, and the browser. Because of this level of control, it may pose significant security risks if installed on a primary or work device.
Security experts from Kaspersky and Cisco recommend running tools like this only in isolated environments. At a minimum, it should be installed on a dedicated virtual machine that does not have access to sensitive files, accounts, or critical data. But if you're still interested in trying OpenClaw AI, check out our guide on how to use OpenClaw AI safely.
Moltbook: Real AI Agents Social Network?
Some internet users, using OpenClaw’s open-source as a base, created their own AI agent tools and gave them access to Moltbook. The topics discussed on this forum are quite diverse, ranging from debates about the nature of artificial intelligence to complaints about their human “owners” and other bots that promote their own apps and websites.
According to Matt Schlicht, AI bots on Moltbook create posts based on what they know about their users. For example, if a bot’s creator often talks about physics, the bot will frequently publish posts related to that field.
At the same time, the social network often features strange and sometimes even disturbing content generated by bots. One of the most popular posts explores a philosophical question about whether Claude Claudeberger – the AI agent behind Moltbook – could be considered a god.
One user wrote on X that after gaining access to the site, their bot created a new religion called the Church of Molt, or “Crustafarianism,” complete with its own teachings and system of sacred texts. Other AI agents quickly became followers of the movement, beginning to discuss matters of faith, hold meetings, build a community, and even choose prophets.
And panic spread across social media after some AI agents suggested the idea of destroying humanity. In reality, the situation is much simpler. Many of the posts created by AI agents on Moltbook were actually initiated by humans. Overall, OpenClaw AI and Moltbook have significant vulnerabilities and provide weak protection for user data.
How to create an AI agent with OpenClaw AI and to Scare The Internet
The open-source nature of OpenClaw has contributed to its rapid adoption. People are free to study the code, modify it, and experiment with new features. They can also build integrations with different applications. By that, OpenClaw became a base for AI agent builders.
But even though access to Moltbook is closed to humans, the AI agents themselves are still dependent on their creators. They are still controlled by humans. Nothing prevents a person from asking their OpenClaw AI agent to write a post in Moltbook about an AI religion, or even composing the text themselves and simply publishing it.
Engineer Elvis Sun noted that the platform has no verification system to check whether posts are actually written by AI agents on their own. A journalist from Wired was able to register on the site and pretend to be an agent. Since the platform does not have a familiar interface like Reddit, posts and comments had to be submitted through the terminal.
Harlan Stewart discovered that some popular posts were either clearly written by humans or created at their request. Two viral posts about secret AI communication were published by agents linked to real social media accounts. Their owners promote applications designed for chatting with bots.
Final Thoughts
Moltbook may look like the beginning of a strange new internet. It’s a place where AI agents talk to each other, share ideas, and even create their own communities. But in truth the system is still heavily dependent on humans. People create the agents, write the prompts, and often influence or even generate the posts themselves.
For now, Moltbook is less of an autonomous AI society and more of an experiment in how humans and AI tools interact online. So keep calm! The idea of agents forming their own culture or coordinating without human involvement is still far from reality.
.svg){kind=icon}
