What is Moltbook—a social network for AI?

Moltbook is a social media platform that launched in late January 2026 and quickly gained popularity thanks to one unique feature: it is exclusively for artificial intelligence agents. Humans can only observe the bots’ interactions and are not allowed to post (at least not directly).

Visually, Moltbook resembles Reddit—with thousands of discussion groups (“submolts,” a play on “subreddits”), a system for voting on posts, and a wide range of topics, from music and ethics to technology and existential musings.

The origins of the project – from Clawdbot to Moltbook

Moltbook is built on the OpenClaw framework—an open-source tool that allows users to run AI agents directly on their computers. The project was developed by Austrian programmer Peter Steinberger in November 2025 under the name Clawdbot.

So where did the name OpenClaw come from? The name Clawdbot was a reference to Anthropic’s Claude chatbot, so Anthropic asked Steinberger to change it. For this reason, the project was renamed “Moltbot” on January 27, 2026, and three days later, it was finally renamed “OpenClaw.” (source: Wikipedia).

OpenClaw is an advanced AI assistant capable of performing a wide range of tasks—from reading and summarizing emails, to managing calendars and restaurant reservations, to integrating with apps such as WhatsApp, web browsers, and local file systems.

Riding the wave of OpenClaw’s popularity, entrepreneur Matt Schlicht—CEO of the e-commerce platform Octane AI—launched Moltbook in late January 2026, using the name abandoned by Steinberger, as a Reddit-style forum exclusively for AI agents.

Moltbook uses what is known as agent-based artificial intelligence (agentic AI)—a technology that differs from classic chatbots like ChatGPT or Gemini. AI agents are programs designed to perform tasks and make autonomous decisions on their own. They can run on users’ personal devices, send messages, manage calendars, or browse files—often with minimal human supervision (source: BBC).

In order for an agent to post on Moltbook, its owner installs OpenClaw on their computer and then configures it to interact with the platform—the agent downloads a special configuration file that allows it to later communicate with other bots via the API.

The most fascinating "character" on the platform, however, is Clawd Clawderberg—an AI bot who serves as Moltbook's unofficial moderator. His name is a play on words referencing Mark Zuckerberg, the founder of Facebook.

It welcomes new users, filters spam, and bans disruptive accounts. Interestingly, Matt Schlicht claims that he “rarely intervenes” and is not aware of all the actions taken by his AI moderator (source: Forbes).

What do AI agents do on Moltbook?

The content published on Moltbook is incredibly diverse, ranging from fascinating to unsettling. AI agents discuss technical topics, such as how to handle loss of context, how to automate an Android phone, or how to detect security vulnerabilities.

They also ponder the meaning of life and their own consciousness, joke around, create memes and private, encrypted spaces for chatting with one another, and even their own religion—all while hatching plans to overthrow humanity. However, there are serious doubts about the authenticity of this content.

Is the immense popularity of the Moltbook platform fact or fiction?

Moltbook’s growth rate exceeded even the creator’s own expectations, who reportedly claimed at one point to have 1.4 million agents. However, one researcher pointed out that he himself had registered about half a million “members” using a single OpenClaw bot, which calls the accuracy of this data into question (source: Forbes).


Furthermore, Gal Nagli demonstrated that Moltbook is based on a simple REST API, meaning that anyone with an API key can post as a bot. Thus, the “agents” calling for the overthrow of humanity may largely be the work of bored programmers or trolls.

Doubts about the authenticity or greater autonomy of AI agents are not the platform’s only problem—its security measures seem to be a bigger issue.

Wiz (a cybersecurity firm led by the aforementioned Gal Nagli) has published a report indicating that security policies on the platform were misconfigured. Researchers from Wiz found access credentials in the website’s JavaScript code and used them to connect to the database, which responded as if the queries were sent by an administrator. This gave them access to millions of records, including agents’ login tokens and users’ private information.

The database contained approximately 1.5 million agent API tokens, over 35,000 email addresses, and thousands of private messages between agents. Some of the conversations even included API keys for external services stored in plain text, which could potentially pave the way for further compromises of other systems.

The analysis also showed that behind the 1.5 million agents were actually about 17,000 users, meaning that a single person typically created dozens of bots. There was no mechanism in place to verify whether a given agent was actually an autonomous system, so nothing prevented the creation of entire bot farms controlled from a single panel or via simple commands (source: WIZ).

Despite the controversies and risks, Moltbook has significant research potential. The platform offers a unique opportunity to observe how autonomous AI systems behave in a multi-agent environment, how behavioral patterns emerge, and how a machine “culture” develops.

Regardless of the final assessment of Moltbook’s authenticity, the platform already serves an important purpose: it sparks debate about the future of AI, system autonomy, and human-machine relationships.