News

Meta Bets on Agent Identity Infrastructure With Moltbook Acquisition

• By John K. Waters
• 03/16/2026

Meta has acquired Moltbook, the agent-native social platform built by Matt Schlicht and Ben Parr, and has folded the founders into Meta Superintelligence Labs under Alexandr Wang. Deal terms were not disclosed. The transaction was expected to close mid-March, with Schlicht and Parr onboarding March 16, per Axios, which first reported the deal.

Moltbook was never a consumer play in any conventional sense. Schlicht launched it in January 2026 as infrastructure for autonomous agents, a place where software, not people, could post, interact, and coordinate. The platform was tightly coupled with OpenClaw, an open agent framework that cycled through the names Clawdbot and Moltbot before its creator, Peter Steinberger, announced in February that he was joining OpenAI and spinning the project into an independent foundation.

What Meta is actually buying is more interesting than the product itself. Meta exec Vishal Shah described Moltbook as having solved core agent plumbing: identity verification, agent-to-agent discovery, and persistent linkage back to human principals. That is the hard part of agentic systems that nobody talks about: not the model, not the UX, but the auth layer. Who is this agent? Who owns it? What is it allowed to do? Moltbook had apparently made a credible first pass at those questions, which likely caught Meta's attention.

The timing fits Meta's broader posture. The company has been scaling Superintelligence Labs aggressively and signaled in January that 2026 would see AI go deeper across both consumer and business products. Owning the identity and coordination layer for agents — especially as those agents begin operating inside business workflows — is a meaningful strategic position, not just a research asset.

There are caveats. Moltbook's reputation was partly built on the spectacle of agents apparently gossiping and self-organizing, but The Verge found evidence of human involvement in some of the most prominent activity on the platform. More damaging: Wiz disclosed last month that a misconfigured Moltbook database had leaked 1.5 million API authentication tokens, 35,000 email addresses, and private inter-agent messages, with read/write access open before the hole was closed. For a company pitching itself as an agent identity infrastructure, that is not a great look.

Existing users can keep accessing the service for now, though both Axios and The Verge indicated that continuity is likely temporary.

The broader read here is straightforward: the model wars are becoming table stakes. The next competitive frontier is the runtime layer — persistence, identity, permissions, and coordination — that enables models to operate as long-lived agents rather than stateless chatbots. OpenAI is hiring toward it. Meta just bought into it. The race to own that stack is now clearly on.