Tech & Science

Meta developing AI agent to shop and act on users’ behalf

Meta is developing a consumer-facing AI agent codenamed 'Hatch' that would autonomously handle tasks like shopping, payments, and document work on behalf o...

Meta is developing a consumer-facing AI agent codenamed “Hatch” that would autonomously handle tasks like shopping, payments, and document work on behalf of users, according to a report from The Information published on May 5. The project represents Meta’s most direct attempt yet to bring the capabilities of open-source agent platforms like OpenClaw to its billions of users in a more accessible package.

Meta developing AI agent to shop and act on users' behalf

From Sandbox to Shopping Cart

The Hatch agent is being trained in simulated web environments modeled after platforms such as DoorDash and Etsy, according to reporting from The Information. Meta aims to complete internal testing by June. The agent is designed to make proactive decisions and retain memory across sessions, allowing it to act without repeated prompting.

Alongside Hatch, Meta is building an agentic shopping tool for Instagram that would let users click on products in Reels or feeds and complete purchases without leaving the app. The company plans to launch the shopping feature before the fourth quarter of this year, positioning it as a direct challenge to TikTok Shop.

Zuckerberg’s Agent Vision

The development follows CEO Mark Zuckerberg’s comments during Meta’s first-quarter earnings call on April 28, where he laid out his ambitions for personal AI agents. “Our goal is not just to deliver Meta AI as an assistant, but to deliver agents that can understand your goals and then work day and night to help you achieve them,” Zuckerberg said. He described OpenClaw — the fast-growing open-source agent that gained widespread popularity after launching in January — as offering “a very exciting glimpse of what types of things should be possible” but “pretty rough” to set up.

Meta plans to eventually power Hatch with its Muse Spark model, the first release from Meta Superintelligence Labs, which launched in April as a multimodal reasoning model with native support for tool use and multi-agent orchestration.

The Competitive Landscape

The move places Meta alongside several major companies racing to deploy autonomous AI agents for consumers. Amazon has its “Buy for Me” agent, while Walmart is building agent-completed shopping within its own platform. OpenClaw, meanwhile, has become a benchmark in the space — an MIT-licensed, open-source system that runs locally and connects through messaging apps to perform tasks ranging from email triage to browser automation. Meta’s advantage lies in its distribution: more than three billion daily active users across its family of apps.

Leave a Reply

Your email address will not be published. Required fields are marked *