Meta's agentic AI called "Muse" has ended its internal testing and officially entered a limited Alpha and Beta closed testing phase. According to Testing Catalog, the app has recently been publicly listed on the App Store platform, and currently, users need to join a waitlist or enter an invitation code to obtain access.

QQ20260908-135931.jpg

As a highly integrated edge-side AI agent, Muse aims to provide intelligent task management services across all device scenarios. In terms of basic connectivity, Muse can directly connect to Gmail, Google Calendar, and health apps such as Peloton and HealthEx; in terms of collaboration and transactions, the agent supports automatic form filling, sending messages on behalf of users, monitoring ticket prices for flights and movies, and making reservations; in terms of security risk control, any payment or key decision requires manual confirmation and approval from the user.

Notably, Muse has deep cross-industry capabilities, allowing direct connection to users' credit cards and bank accounts for real-time spending tracking, and it can intelligently suggest canceling low-frequency redundant subscriptions based on the user's performance history to save money.

Previously, Meta had launched the Muse Spark large model system for Gemini and localized AI models. The launch of Muse marks Meta's accelerated transition from pure large language model development to personal AI agents with high execution capabilities and long task chain processing abilities.

As agent technology becomes deeply embedded in sensitive scenarios such as personal privacy and finance, how it balances efficient service with user privacy and trust will become the key to whether the product can achieve large-scale commercialization.