Internal documents at Meta indicate plans to debut a consumer-focused AI agent codenamed Hatch within the coming weeks, alongside a target launch for a new model, dubbed Watermelon, in October. This product aims to rival the open-source agent framework OpenClaw, with internal pricing discussions touching on a premium tier of US$199.99 per month, seeking revenue beyond advertising to fund substantial AI infrastructure investments. A Meta spokesperson declined to comment on these specific plans.

The product format and launch window for Hatch are based on internal documentation, which outlines the rollout for the consumer version of the open-source agent, OpenClaw, in the coming weeks. Concurrently, the company is targeting an October release for the new Watermelon model. The documents peg the launch timeframe for Hatch for late August or early September. It remains unclear whether Watermelon will be integrated into the Muse series or offered as a standalone release.

Early prototypes of Hatch have been trained to interact with websites like DoorDash, Etsy, Reddit, Yelp, and Outlook. The intended interface is a customizable dashboard displaying agent-generated tools or skills, such as fitness tracking and travel itineraries. Reports indicate Meta has discussed tiered subscription models, with the highest tier potentially reaching US$199.99 per month, offering increased usage limits. In a parallel distribution effort, Meta is preparing to launch a platform on WhatsApp, allowing users to integrate other AI agents into their conversations, with a small-scale rollout targeted for this week.

The parallel development of Hatch and the WhatsApp channel suggests the company is aiming to embed agent capabilities into existing messaging entry points rather than creating a standalone app. OpenClaw itself is an MIT-licensed open-source personal agent initiated by developer Peter Steinberger and later supported by OpenAI. It runs a Gateway on users' own devices, executing tasks via channels like WhatsApp, Slack, and Discord, emphasizing local control and self-supplied model keys. By turning similar capabilities into a managed consumer product, Meta is connecting proven open-source interaction methods to its own account system and app distribution.

Behind Hatch is a recent acceleration in releases from Meta Superintelligence Labs. April saw the debut of the first Muse Spark, focused on multimodal reasoning, tool use, and multi-agent collaboration. This was followed by an upgrade to Muse Spark 1.1 in July, enhancing coding and computer operation. On August 5th, Muse Spark 1.2 and the terminal coding agent Muse Code were released. In official evaluations, version 1.2 achieved 82.9% on Terminal-Bench 2.1 and 59.3% on DeepSWE v1.1, up from 76.2% and 53.0% for version 1.1, respectively. The standard API pricing is US$1.25 per million input tokens, US$0.15 for cached input, and US$4.25 for output. On August 10th, the company also open-sourced Muse Glimmer, a roughly 30-billion parameter model under the Apache 2.0 license, aimed at single-GPU local agents and coding workloads, with a promise to release Muse Spark 1.2 weights later.

Mark Zuckerberg has simultaneously written about expanding open-source access. The product portfolio now consists of a closed frontier model and an open-source flank: Hatch targets managed agents and subscriptions, while Glimmer and open-source weights maintain the developer ecosystem. If Watermelon launches in October, it would represent the next flagship iteration following Muse Spark. The documents do not specify parameter counts, whether the weights will be public, or if it will directly power Hatch. The timeline remains subject to change.

The US$199.99 price point places Hatch at the top tier of consumer AI subscriptions. Mainstream plans like ChatGPT Plus, Claude Pro, and Gemini AI Pro are around US$20 per month. However, premium options such as ChatGPT Pro, Claude Max 20x, and Google AI Ultra all reach approximately US$200 per month, offering higher usage and more powerful models. Meta's premium discussion for Hatch aligns its price anchor with OpenAI, Anthropic, and Google. SemiAnalysis estimates that heavy users on Anthropic's US$200 monthly plan can consume about US$8,000 worth of equivalent API usage, while similar OpenAI plans can reach up to US$14,000, a gap largely driven by token consumption from agentic coding tasks. For vendors, top-tier subscriptions are a means of customer acquisition and lock-in, but on the cost side, they represent a subsidy. If Hatch focuses on tasks like food delivery, shopping, email, and local services, its token path differs from pure chat, and gross margins will depend on the depth of invocation and whether traffic is redirected to ads or e-commerce commissions.

Meta's unique asset is its distribution network. Facebook, Instagram, and WhatsApp collectively serve billions of daily active users. WhatsApp business messaging contributed approximately US$1 billion to "Other Revenue" in the second quarter of 2026, a 73% year-over-year increase. If agents can complete food orders, customer service, and transactions within WhatsApp, subscriptions become just one layer of pricing; transactions and advertising could represent a larger monetization surface. The market context is clear: high capital expenditures require visible product financials. In Q2, Meta reported revenue of US$60.8 billion, up 28% year-over-year, with advertising revenue at US$59.36 billion. Capital expenditures, including finance lease principal, were approximately US$31.08 billion, and free cash flow fell to US$784 million, a 91% decrease year-over-year. The full-year capital expenditure guidance was narrowed to between US$130 billion and US$145 billion. The company ended the quarter with US$90.26 billion in cash and marketable securities and US$83.66 billion in long-term debt. Advertising still accounts for about 98% of revenue. Hatch is seen in internal documents as part of Zuckerberg's push to monetize AI investments and diversify beyond advertising.

The risks are equally concrete. The spokesperson declined to comment, and internal timelines are subject to slippage. OpenAI and Anthropic have already established positions in the agent and coding space, while Google is integrating agent capabilities into Workspace and Search. Hatch must prove that placing orders, checking reviews, and writing itineraries within existing apps can achieve completion rates and security sufficient to justify top-tier pricing. Regulatory matters, such as child safety litigation, remain ongoing. If consumer agents expand automated operations on external websites, compliance and account security costs will enter the expense column. In the short term, the valuation impact is limited; an unconfirmed internal timeline is unlikely to alter a spending framework of US$130 billion. In the medium term, if Hatch launches on schedule and shows observable subscription or transaction revenue, the market will begin converting the narrative of "AI investment accelerating core business" into line-item numbers. Until then, the more important signals to track are whether Watermelon materializes in October, the scale of the WhatsApp agent beta, and whether management includes consumer agents in formal product guidance during the Q3 earnings call.