Meta bets on AI agents: develops its own "Hatch," Instagram shopping tool targets TikTok

Meta is fully betting on AI agent technology, viewing it as the next-generation core product form and synchronously advancing two key initiatives internally. On May 5, technology media The Information reported that, according to informed sources, Meta is developing a consumer-facing AI agent product, internally codenamed “Hatch”, aiming to complete internal testing by the end of June. The company also plans to integrate an independent AI shopping agent tool into Instagram and strives to launch it before the fourth quarter of this year. The report says these plans directly reflect CEO Mark Zuckerberg’s drive for an AI agent strategy. Last week, on the company’s quarterly earnings call, Zuckerberg stated that Meta’s goal is “to build agents that understand users’ objectives and help them achieve those objectives around the clock.” Meanwhile, Meta announced it would increase its AI infrastructure capital expenditures this year to a maximum of $145 billion. The launch of Instagram’s shopping agent is viewed internally as a crucial move to compete head-to-head with TikTok Shop. Hatch: From OpenClaw to In-house Agent According to reports, Hatch was directly inspired by OpenClaw—a tool that quickly became popular in tech circles for its ability to build autonomous AI agents. Zuckerberg once attempted to acquire OpenClaw, but according to its founder Peter Steinberg in a podcast, the tool was eventually acquired by OpenAI in February this year, and Meta’s attempt failed. Against this backdrop, Meta has turned to independently developing a consumer-grade agent with similar functions. As reported, informed sources said Hatch is currently powered by Anthropic’s Claude Opus 4.6 and Claude Sonnet 4.6 models, but will switch to Meta’s in-house AI model Muse Spark upon official launch. To test Hatch’s actual capabilities, Meta set up a “sandbox” network environment simulating real websites like DoorDash, Etsy, Reddit, Yelp, and Outlook, for agents to train and be tested under controlled conditions. Zuckerberg also acknowledged the technical challenges of rolling out OpenClaw-like tools to Meta’s billions of users during the earnings call, including heavy reliance on large-scale infrastructure and ensuring the product is simple and easy to use—OpenClaw remains too complex for most non-technical users. Capability Iteration: Initiative, Memory, and Tool Invocation Meta is currently accelerating the improvement of Hatch’s core abilities on multiple dimensions. Reportedly, informed sources say the company is focusing on enhancing Hatch’s proactive decision-making capabilities, enabling it to take actions independently at appropriate times instead of passively waiting for user commands. At the same time, Meta is expanding the model’s context processing window and strengthening its cross-conversation memory capabilities so the agent can continuously remember relevant user information across different sessions. The company is also optimizing how agents respond and their logic in selecting and invoking external tools. Notably, Meta internally already has an AI agent, MyClaw, for employee use, which accesses work files, summarizes company forum posts, and provides technical advice. However, according to previous reports by The Information, MyClaw recently triggered a major security alert inside Meta—a staff member followed an erroneous recommendation from the agent, leading to sensitive company and user data being accessed by unauthorized employees. This incident highlights the ongoing real-world challenges in AI agent reliability. Instagram Shopping Agent: Targeting TikTok Shop On the consumer side, Meta’s other focus is deeply embedding an AI shopping agent into Instagram. According to reports, informed sources say the tool’s core functions include: users can directly click products in Instagram Reels or feed, view detailed information, jump to external webpages, and complete purchases within the platform, all without leaving the app. This feature builds on the upgraded AI shopping experience launched by Meta in March, which includes more detailed product information presented with AI and a new checkout process allowing users to purchase directly through clicking ads. Informed sources also revealed that Meta hopes this tool will intensify the competitive pressure on TikTok Shop. TikTok Shop, leveraging deep integration of short videos and e-commerce, has established significant advantages in social e-commerce, while Meta aims to counter this with Instagram’s large user base and AI capabilities. In a broader competitive landscape, Google launched Gemini Enterprise for Customer Experience in January, which includes AI shopping agent features, supporting product recommendations, cart building, and auto-checkout after user authorization; Amazon offers an AI shopping assistant, Rufus, helping users track prices, research, and buy products. Analysts point out that the advancement of Hatch and Instagram’s shopping agent reflect Meta’s large-scale capital investment in AI seeking commercial returns. Zuckerberg positions AI agents as the core of the “Personal Super Intelligence” vision and describes them as all-day tools for helping users achieve their goals during the earnings call. As Meta raises its AI infrastructure capital expenditure cap to $145 billion this year, market attention on when its AI investments will translate into substantial revenue continues to heat up. Hatch's ability to strike a balance between technical complexity and mass accessibility, and whether the Instagram shopping agent can effectively impact the social e-commerce market, will be key indicators for evaluating this strategy. Risk warning and disclaimer The market is risky, and investment must be cautious. This article does not constitute individual investment advice, nor does it take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, viewpoints, or conclusions in this article fit their specific circumstances. Investment based on this is at your own risk.