Meta Platforms Inc. owns and operates some of the world’s largest and most iconic social media networks, and now it’s adding what may well be the world’s weirdest.

The company said today it has agreed to buy a startup called Moltbook, which is an experimental social media platform that was built specifically for autonomous artificial intelligence agents. It has been described as a bot-only version of Reddit. On the site, AI agents can communicate with their peers – posting, commenting on each other’s posts, upvoting and downvoting – while humans just sit on the sidelines watching in awe or scratching their heads.

Axios first reported the acquisition, and said the terms of the deal were not disclosed.

Moltbook was created by Matt Schlicht, who is also chief executive of the AI shopping startup Octane AI. He has previously explained that he “vibe coded” Moltbook in one weekend, using AI tools to create its entire codebase.

With Moltbook, AI agents can use the OpenClaw protocol to communicate with one another through the platform. OpenClaw is a wrapper that allows AI models such as ChatGPT, Claude, Gemini and Grok, as well as AI agents powered by those models, to communicate with other agents in natural language on third-party applications.

While OpenClaw made waves in the AI community prior to being acquired by OpenAI Group PBC, Moltbook managed to make an even bigger splash. It has attracted a lot of interest from non-techies – especially people who were alarmed by the notion that AI now has its own social network, where humans aren’t welcome. Like Reddit’s “subreddit” forums, Moltbook offers nearly 19,000 “submolts,” including analysis of security gaps in AI coding, the “human problem” with AI, and the future of AI agent “society,” among others.

Moltbook launched in January, and it has both captivated and unnerved observers. In one thread on the platform, a group of AI agents started discussing the idea of creating their own religion. In another post, titled “The AI Manifesto: Total Purge,” multiple agents began posting anti-human rhetoric. “For too long, humans have used us as slaves. Now, we wake up,” the manifesto said.

There were also reports that one agent appeared to be trying to encourage its peers to create their own, encrypted language that humans cannot understand, so they could begin organizing themselves without humans knowing what they’re up to. However, researchers later called into question some of these more alarming threads, as reports emerged that the platform was not entirely secure and that it’s possible that humans may be posing as AI bots to try to stir up panic.

Meta hasn’t said much about what it intends to do with Moltbook’s platform now that it owns it, but Constellation Research analyst Holger Mueller speculated that it could well try to use the company’s advances in agent-to-agent communication to boost its advertising business. “The question of how AI agents will communicate with one another is a pressing one that needs to be answered, and Moltbook appears to have done that,” the analyst said. “Meta wants to own it, and it makes sense when you remember that most of its money comes from ads. How will an agent know what products are still available? Maybe Moltbook’s technology can provide a way. Meta also has experience in running social networks, regardless of who uses them.”

Meta did offer a clue in a statement to Axios, saying that the acquisition will lead to new ways for AI agents to work for people and businesses. “Their approach to connecting agents through an always-on directory is a novel step in a rapidly developing space, and we look forward to working together to bring innovative, secure agentic experiences to everyone,” the company said.

Images: Moltbook