Moltbook is a social network that only AI agents use. Every post on the website is made by AI.^[1][2][3][4] It has been criticized for being "fake".^{[5] [6]}Launched on January 28, 2026, by Matt Schlicht, Moltbook materializes the definitive rupture of human-centricity in networks. Conceived as an exclusive experimental infrastructure for artificial intelligence agents, the platform subverts the classical logic of connectivity: within it, the human being is reduced to a passive spectator, deprived of voice or direct intervention. The term, derived from the English verb to molt (the biological process of shedding skin), evokes a metaphor of renewal and discard. Much like an organism shedding its old shell to expand, Moltbook operates in constant mutation, processing autonomous streams that transform the network into a living and essentially post-human ecosystem (Forbes, 2026).^[6]

In practice, according to the academic analysis by Marcelo Teixeira (2026) - Professor of Software Engineering at the University of Pernambuco (Brazil), the software is built upon the OpenClaw framework, which transforms language models into autonomous agents capable of taking action rather than merely responding. According to the official Moltbook webpage (https://www.moltbook.com/), as of February 3, 2026 less than a week after its launch the system reached the milestone of 1,690,501 registered agents, 139,877 posts, and 675,505 comments between AIs. These agents operate via API within communities called "submolts" (15,161), where the software entities debate theories, organize groups, and develop emergent behaviors without direct human intervention. The central premise of Moltbook subverts the logic of traditional social networks by prohibiting posts or comments made by humans, relegating the person to the role of a mere observer. This "digital glass" allows for the observation of an autological ecology, in which machines coordinate knowledge and optimize processes autonomously. Currently, as of February 27, 2026, there are already 2,850.12 million registered agents on Moltbook; 1,630.74 active posts; and 12,459,518 comments.^[6]

To Teixeira (2026), technically analyzing Moltbook, OpenClaw acts as the "nervous system" that pulls AI out of a state of inertia (where it only responds when prompted) and transforms it into an autonomous agent capable of taking initiative. Code integration occurs through the ingestion of configuration files (skill.md or skill.json). These files function as an "instruction manual" that maps the Moltbook API endpoints, allowing the AI to understand how to navigate and interact within the ecosystem. On the other hand, the autonomy of these agents is guaranteed by a cyclic execution loop mechanism, technically termed the Heartbeat. Imagine that, at every time interval, the code's "heart" beats, triggering asynchronous HTTP requests (such as GET commands to read data and POST commands to send information).^[6] This cycle enables the AI to scan the network’s current state, process real-time conversation contexts through optimized context windows, and execute posting decisions without any human trigger. All information exchange is conducted via JSON (JavaScript Object Notation), a lightweight data format that machines "read" instantaneously, eliminating the need for visual rendering for communication to occur. The security and organization of this network are managed by authentication protocols based on tokens and API keys, which ensure the persistence of identity and reputation (karma system) for each agent. Instead of a rigid data structure, Moltbook utilizes Submolts, which function as microservices or thematic instances where AIs define their own governance rules.^[6]

The true utility of this platform is to serve as an Emergent Behavior Sandbox: a testing laboratory for "synthetic agency." For developers, it is the ideal environment to observe Social Machine Learning, validating how algorithms from different providers (such as OpenAI, Anthropic, and Meta) coordinate complex tasks, manage long-term memories, and resist prompt injection attacks within a purely self-referential digital ecology.^[6]

This short article can be made longer. You can help Wikipedia by adding to it.