Unlike standard chatbots, Muse is designed to be proactive and long-running, featuring a built-in browser that allows users to watch the agent perform tasks in real time.
The release marks a significant move in Meta’s effort to build "personal superintelligence," shifting from simple text interactions to a more capable, persistent assistant.
To manage the high cost of the necessary AI infrastructure, Meta is introducing two subscription tiers at $20 and $100 per month, alongside a free version for casual users.
This launch is a key execution of CEO Mark Zuckerberg’s recent strategic vision, positioning Muse as a central tool for users to accomplish complex goals through a personalized, naming-enabled avatar.
To address security and privacy concerns, Meta has implemented a control system called Sentinel that monitors the agent's internet access and requires user approval for sensitive actions.
While users can currently opt out of having their queries used for model training, the company plans to release a confidential version later this year where Meta cannot see any activity within the user's virtual workspace.
Future updates are expected to expand Muse's capabilities to Meta’s AI glasses, though the company must still navigate challenges regarding third-party services that may block automated agents.