Meta Launches Muse Connectors Platform to Challenge Apple's Ecosystem
Meta officially launched the "Muse Connectors" platform on September 18, allowing third-party developers to integrate their services' APIs with Meta's personal assistant, Muse, just 10 days after Muse's official launch on September 8.
Muse was launched in the U.S. market on September 8 and can be accessed via web (muse.ai), iOS, Android apps, and WhatsApp, with plans to integrate with Meta's smart glasses. Pricing includes a limited free version, a "Power" version at $20 per month, and a "Maximum" version at $100 per month, all requiring a linked payment card. Its capabilities include sending and receiving emails, booking travel, reducing bills, filling out forms, planning, converting recipes into shopping lists, sending invitations, and completing purchases, with checkout integrated with Stripe's Link payment tool, and Shop Pay and 1Password integrations in development. Within 10 days of launch, Muse topped the U.S. iPhone free app chart, surpassing competitors like ChatGPT, Gemini, and Claude.
The "Muse Connectors" platform operates by allowing third-party developers to provide APIs, while Muse handles the conversation interface, user context understanding, and autonomous execution—Zuckerberg summarized it as "You bring the API." After users express their needs in natural language, Muse determines intent, selects the appropriate connector, and orchestrates execution steps, capable of browsing the web, filling out forms, connecting applications, completing purchases, and continuing tasks after the user closes the app. Unlike traditional app stores that rely on users to actively search and download, the sorting and recommendations of connectors are entirely controlled by Meta, referred to by industry observers as "the search ad space of the agent era."
In terms of security mechanisms, Muse provides each instance with an independent isolated virtual machine (Muse Secure VM) and is audited by an independent Sentinel system before actions are executed. Users can set read and write permissions for each connected service, with sensitive operations requiring additional confirmation, and all actions are logged for auditing; Meta stated that related conversation data will not be used for advertising systems, and Muse itself cannot access user passwords or payment information.
Alongside the Muse platform, Meta is also advancing its next-generation large model codenamed "Watermelon"—according to Alexandr Wang, head of Meta's "Super Intelligence Lab," this model has reportedly matched the performance of OpenAI's GPT-5.5 in internal benchmark tests, though specific testing details have not been disclosed, and the external response to this claim remains cautious. Watermelon is an upgrade based on the "Muse Spark" (internal codename "Avocado") released in April this year, reportedly requiring computational power exceeding that of its predecessor, with no official release timeline yet.
From the perspective of industry structure and profit distribution, the core target of the Muse Connectors platform is the distribution rights of traditional app stores: Apple App Store and Google Play have long relied on users actively searching and downloading, with developers needing to obtain rankings and exposure themselves, while Meta intercepts traffic directly at the "willingness level" of user demand through Muse, deciding which third-party API to call to complete tasks, effectively shifting the entry and ranking rights of app distribution from app stores to the agents themselves. This means that developers may no longer need to maintain complete independent apps in the future, only needing to expose API interfaces to connect with Muse and gain users, potentially weakening the traditional app model and the advertising revenue model of app stores. However, this impact currently remains more at the potential stage—Muse has only been online for 10 days, and its operation still heavily relies on the permissions and computational support of the iOS and Android operating systems, with the actual substitution effect on the Apple and Google ecosystems depending on whether the user base can continue to expand and the richness of the connector ecosystem. Additionally, Meta's aggressive push comes just after the company paid $18 billion in a settlement for a class action lawsuit regarding minors' online safety, making the balance between functional expansion and user trust restoration a key variable for the platform's successful implementation.
Source: Public Information
ABAB AI Insight
Meta has consistently adopted a strategy of "rapid high-frequency releases and quick iterations to bind the ecosystem" for its AI products—from the initial release of "Muse Spark" (internal codename Avocado) in April, to the official version of Muse on September 8, and then the launch of the connectors platform on September 18, marking two significant updates within 10 days, continuing Zuckerberg's historical approach of intensive iteration and seizing user mindshare in projects like VR/metaverse and the Llama open-source model.
In terms of capital and resource investment paths, the computational power required for Watermelon is reportedly an order of magnitude greater than that of its predecessor, indicating that Meta continues to invest heavily in AI infrastructure and computational power; at the same time, Muse adopts a "free traffic + subscription fee ($20/$100 monthly)" business model, attempting to rapidly scale user numbers in the short term with a free version, and then commercialize through subscription fees and potentially a connector revenue-sharing mechanism, which is a typical capital path of first investing in computational power to gain scale, and then using scale to achieve commercialization.
This model can be compared to the historical process when Apple App Store and Google Play established application distribution rules—at that time, Apple and Google leveraged their entry advantages at the operating system level to set application sorting, review, and revenue-sharing rules, reshaping the entire distribution landscape of the mobile internet; the Muse Connectors platform aims to replicate similar entry control at the AI agent level. In terms of the industry's current stage, AI agents are in the early expansion phase transitioning from "single-point AI applications" to "comprehensive task execution and distribution platforms," with Muse topping the iPhone free chart and the launch of the connectors platform being key actions in this phase to compete for entry positions.
Structural judgment: This event represents a restructuring of the industry chain. If the Muse Connectors platform truly operates effectively, it means that the application distribution segment previously dominated by the Apple App Store and Google Play may be partially transferred to Meta's agent entry—developers will no longer need to maintain complete independent apps and compete for app store rankings, but only need to provide APIs to connect with Muse to reach users, which will redefine the power boundaries of the "application—operating system—distribution platform" industry chain in the mobile internet era, representing a typical case of industry chain restructuring.
ABAB News · Cognitive Laws
Apps will disappear, APIs will not.
Who intercepts willingness, who defines entry.
If computational power is insufficient, agents will lead the way.