Meta readies OpenClaw rival 'Hatch' and a new model this fall
Two reports out of The Information sketch Meta's next moves in the agent race: a personal AI assistant codenamed Hatch, reportedly built to compete with the open-source agent harness OpenClaw, is scheduled to land in late August or early September, with a follow-up flagship model dubbed Watermelon due in October. Neither has been officially confirmed by the company, and timings reportedly come from internal documents.
Meta is reportedly days away from shipping Hatch, its answer to the open-source agent OpenClaw, with a flagship Watermelon model to follow in October. The agent platform — described in internal docs, per The Information's Jyoti Mann — would give Meta a consumer-oriented rival to the open-source harness that set the personal-agent world alight this year. Where OpenClaw spread bottom-up through developer communities and GitHub stars, Hatch is a first-party, productized bet: the same reason Meta has been funneling effort into Muse and building agent tools across WhatsApp, Messenger and Instagram. A launch inside the next two weeks would put Meta first to convert the OpenClaw phenomenon into a mainstream, packaged product for its roughly three-billion-user base.
The Watermelon model, slated for October, reads as the intelligence layer under those agent ambitions. Meta has kept its frontier models quiet while rivals iterate, and a fall flagship would restock its agent stack with a homegrown brain rather than renting one. The timing matters too: landing Hatch before the summer ends and a model to power it a month later is a deliberate two-step — get the surface product out, then upgrade the engine underneath. Both are reportedly unconfirmed, and Meta's historical launches have slipped before, so the dates carry real uncertainty.
What's striking is that this echoes a pattern the whole industry is converging on. Rivals are racing proprietary agent platforms against the same open-source wave Meta is now cloning — Microsoft's Agent Lightning, OpenAI pushing agents into Kiro, and the run of open harnesses all point the same direction. Meta's edge, if the reports hold, is distribution: it can slap Hatch in front of a user base no agent startup can match. The open question is whether a cloned OpenClaw with Meta's reach actually outruns the community that made OpenClaw worth cloning in the first place.
What to watch: confirmation from Meta, and whether Hatch ships with a real agentic hook or is a thin re-skin of what's already public in the open-source harnesses.
Will a productized, three-billion-user clone of an open-source agent actually win the personal-agent race — or will the community's speed prove more valuable than Meta's reach? Tell us in the comments.
Sources: Techmeme / The Information
