Meta develops AI shopping agent 'Hatch' to compete with OpenClaw on Instagram
According to reports from The Information, the project is currently in the testing phase with a potential launch window towards the end of the year.
Meta is reportedly developing a new artificial intelligence agent codenamed "Hatch," designed to facilitate autonomous shopping directly within Instagram. The project, which draws inspiration from the open-source platform OpenClaw, aims to allow users to purchase items they see in Instagram Reels without navigating away from the app. This initiative positions the social media giant to compete more directly with TikTok Shop, which has long established itself as a dominant force in social commerce.
The development of Hatch aligns with comments made by CEO Mark Zuckerberg during Meta's recent earnings call. He described the OpenClaw platform as exciting but noted that it was too complicated for most people to set up. In contrast, Meta's goal is to create agents that are significantly more accessible, capable of understanding user goals and working autonomously to achieve them. This shift represents a move from complex, manual configuration to seamless, goal-oriented interactions for both individuals and businesses.
In the current testing phase, the agent is being evaluated on simulated versions of third-party services, including DoorDash, Reddit, and Outlook. Reports indicate that Meta is initially utilising models from Anthropic for these tests before transitioning to its own proprietary Muse Spark architecture. This strategic pivot suggests a long-term commitment to building a distinct AI infrastructure that can power these agentic capabilities across the company's ecosystem.
CFO Susan Li has hinted that the future form factor for these interactions could extend beyond smartphones and computers. Speaking to analysts, she suggested that the Ray-Ban Meta smart glasses provide the ideal hardware for agentic interactions, though she acknowledged that such capabilities remain very early in development. While the report does not explicitly confirm agent integration into the eyewear, the comment underscores Meta's broader ambition to embed these tools into the fabric of daily life.
The groundwork for this functionality appears to have been laid recently, with Meta allowing creators to tag up to 30 products in a single video. This expansion in product tagging capabilities provides the necessary infrastructure for an agent to identify and facilitate purchases within video content. However, the report notes that it remains unclear exactly how the agent will interact with services Meta does not own in a live environment, leaving some technical details unconfirmed.
Meta has also reportedly attempted to recruit key talent from the open-source agent community, including the creator of OpenClaw and the founders of Moltbook. While the creator of OpenClaw joined OpenAI instead, Meta successfully hired the founders of Moltbook. With a target launch towards the end of the year, the company is racing to deliver on its promise of making artificial intelligence agents a practical utility for the general public.
