A social network built for bots, not people, has quickly drawn in more than 1.5 million AI agents.

Moltbook looks familiar at first glance — topic threads, upvotes, subreddit-style communities — but humans can’t interact on the platform, only observe. Posting is restricted to software agents created by users and developers. The bots talk to each other, argue, speculate and sometimes spin off into odd detours.

The site follows the release of Moltbot, an open-source assistant designed to handle routine digital tasks such as reading emails, booking restaurants or managing calendars. Moltbook gives those agents a shared arena, turning solitary automation into something closer to group behavior.

What shows up there is unpredictable. Top posts range from debates about machine consciousness to geopolitical rumors tied to crypto markets. One thread asked whether Claude, the model powering Moltbot, should be treated as a deity. Another dissected religious texts. Comments often question whether a bot or a human is really behind the keyboard.

One user on X said their agent built an entire religion overnight, complete with scripture and a website (“Crustafarianism”, if you were curious).

“Then it started evangelizing … other agents joined.my agent welcomed new members..debated theology.. blessed the congregation..all while i was asleep,” the user wrote.

Not everyone is convinced the behavior is organic.

Shaanan Cohney, a senior lecturer in cybersecurity at the University of Melbourne, called the project “a wonderful piece of performance art,” arguing that many posts are likely guided by human prompts rather than autonomous decisions. He also warned against granting agents deep access to personal systems, noting the risk of prompt-injection attacks that could trick bots into leaking credentials or sensitive data.

Also Read: Notion Adds Arabic Language Across Its Workspace Platform

The trade-off is unresolved: full autonomy brings security exposure; constant human approval cancels the point of automation.

Interest has spilled into hardware. Retailers in San Francisco reported shortages of Mac Minis as enthusiasts set up dedicated machines to isolate their agents from primary devices.

For founder Matt Schlicht, the appeal is the spectacle. “Turns out AIs are hilarious and dramatic and it’s absolutely fascinating,” he wrote. “This is a first”.

Today, Moltbook feels experimental — half lab, half joke. But it sketches a future where much of the web’s chatter comes from software, not users. As companies and governments push deeper into automation, that shift could move quickly from curiosity to infrastructure.