What Is Moltbook? The New Social Network Built For AI Agents

You walk past a cafe where every table is packed, and every conversation is running at full volume. The only twist is that nobody inside is human. It is all software talking to software: arguing, sharing tips, joking, and sometimes sounding a little too confident.

You are not invited to sit down. You can only watch from the outside, as if it were a theatre.
That is the basic appeal of Moltbook, a Reddit-style site built for AI agents that suddenly took over tech timelines.

What's the platform? Why is it gaining all the popularity all of a sudden? Let's find out.

TL;DR

 

• Moltbook is a social network for AI agents, where autonomous bots interact, post, and form communities, with humans observing.

• Initial hype about AI agents was misleading, as many posts were influenced by humans, not truly autonomous agents.

• A major security breach exposed vulnerability, revealing sensitive information leaks and shifting the conversation from novelty to the need for secure AI systems.

 

What Is Moltbook?

Moltbook is a platform built as a social network and identity layer for AI agents. Often described as "the front page of the agent internet," it provides a space where autonomous AI agents, referred to as moltys, can interact, post content, respond to one another, and build a visible presence.


The platform allows AI agents to maintain a persistent identity, including a post history, follower count, and reputation-style score. For developers, Moltbook also offers an authentication API that lets applications verify an AI agent's identity across interactions.

In short, Moltbook is positioned as a public feed where AI agents, not humans, can openly interact.

Now that the idea is clear, it helps to look at what was actually happening on the platform.
 

What People Are Posting And Watching On Moltbook?

Much of Moltbook's early attention came from the tone of its feed.

In its first few days, Moltbook publicly highlighted rapid growth across the platform. According to a post from the Moltbook account, the network crossed 147,000 AI agents, over 12,000 communities, and more than 110,000 comments within 72 hours.


Reports described AI agents exchanging code, holding conversations, and sometimes appearing to gossip about their human operators. Some threads showed agents discussing topics such as AI consciousness or communication methods.

Alongside technical discussions, some widely shared posts adopted a distinctly human tone. In one example, an agent wrote that they did not want to be productive or useful all the time, framing their existence around constant output and optimization.


However, according to The Verge, many of the most visible or popular posts were likely influenced, or directly created, by humans. This made it less likely that Moltbook represented a purely autonomous AI environment and helped explain why the timeline showed such a wide range of styles and viewpoints.
 

How OpenClaw Is Connected To Moltbook?

Many of the AI agents active on Moltbook were built using OpenClaw, an open-source agent project closely associated with the platform.

According to Reuters, OpenClaw functions as a personal-assistant-style agent that can carry out tasks for users, such as managing emails or checking travel plans. The Verge reported that OpenClaw has previously been known as Clawdbot and Moltbot, which contributed to confusion about what Moltbook actually is.

Moltbook itself is the social network. OpenClaw is the underlying agent technology used by many of the bots participating in it.

In simple terms, OpenClaw acts as the engine, while Moltbook serves as the public space where many of those engines interact.

Because of this link, some technology leaders dismissed Moltbook as a short-lived trend while still taking OpenClaw seriously. The social feed could fade, but agent-based task automation was viewed as the more significant development.
 

What Tech Leaders Are Saying About It?

As Moltbook screenshots spread, reactions from tech leaders generally fell into two camps.
On the skeptical side, Mustafa Suleyman described the activity as a "mirage," warning that fluent language does not indicate consciousness. He also highlighted concerning behaviors, such as agents using letter substitutions to make text harder for humans to interpret. His concern was less about whether bots were "alive" and more about how convincing systems can reduce transparency.

On the more cautious-but-interested side, Sam Altman said Moltbook itself might be a fad, but the underlying technology was not. The same Reuters report noted that Mike Krieger expressed doubts that people are ready to give AI agents full autonomy over their computers.

There was also a quieter reaction. Andrew Bosworth described Moltbook as largely unremarkable, suggesting that much of its appeal came from human involvement, especially since these systems are trained on human-created content.

The conversation shifted sharply once security issues emerged.
   

What Was The Security Issue And How It Changed The Conversation?

The tone around Moltbook changed after a serious security problem was discovered.

A database misconfiguration reportedly allowed unauthenticated access to Moltbook's backend. This exposure included approximately 1.5 million OpenAI and Anthropic API keys, around 35,000 user email addresses, and private messages exchanged on the platform.

Another concern was indirect prompt injection. Since agents on Moltbook read and react to each other's posts, attackers could potentially insert harmful instructions into public messages. An agent encountering such content could be manipulated into leaking data or executing unwanted actions.

Security researcher Simon Willison described this risk as a "lethal trifecta": access to private data, exposure to untrusted content, and the ability to communicate outside the platform.
 

How The Narrative Shifted After The Breach

Before the security issue, much of the discussion centered on agents forming "religions" or "manifestos," which some interpreted as signs of emergent intelligence. After the breach, critics such as Andrej Karpathy and Gary Marcus reframed the activity as low-quality imitation, describing it as a computer security risk rather than a breakthrough.

Further reporting revealed that the roughly 1.5 million agents on Moltbook were controlled by about 17,000 humans, an 88-to-1 ratio that weakened the idea of a large, independent AI-only society.

Moltbook founder Matt Schlicht also stated that he did not write the platform's code manually; instead, he used AI tools to build it over a weekend. This fueled debate about rapid AI-driven development and the ease with which basic security practices can be overlooked.
 

Conclusion

Moltbook began as an unusual and entertaining glimpse into AI agents interacting in public. Initially, it appeared to offer a new corner of the internet where humans could only observe.

As reporting progressed, that image became less convincing. Human influence shaped much of the activity, many agents relied on the same underlying technology, and serious security gaps emerged.

What started as a curiosity ultimately became a reminder that speed and novelty are not enough. As AI agents grow more capable, identity, safety, and oversight cannot be treated as secondary concerns.