Meta has introduced Muse, a personal artificial intelligence agent designed to go beyond answering questions by taking action on behalf of users. The company says Muse can manage tasks, support projects and turn long-term goals into personalised action plans.
Muse is powered by Muse Spark, which Meta describes as its most capable AI model to date for real-world agentic tasks. The system can perform activities such as sending emails, booking travel, filling out online forms and making purchases, while also working on longer-term tasks after a user closes the app.
The agent can use a dedicated browser to interact with online services and can return to users when a task requires approval or when circumstances change. Meta said Muse can also help users achieve broader objectives by coordinating their time and resources and adjusting plans as their needs evolve.
Muse is designed to learn from conversations and remember information that users consider important. For example, it can turn a recipe saved on Instagram into a grocery list, recommend a menu for an event and take dietary requirements into account when helping organise invitations.
Meta said privacy and security were central to the development of Muse. The agent operates within Muse Secure VM, a dedicated virtual machine in the cloud that stores the agent, user data and credentials separately from other agents.
A separate security system called Sentinel controls Muse’s access to the internet and must approve online activity. Muse does not directly see users’ passwords or payment details, while sensitive actions such as sending emails or making purchases require user approval. Users can also control which applications Muse can access and the level of permissions it receives.
For online purchases, Muse can use Link, Stripe’s digital wallet, with a one-time-use card designed to keep users’ actual card details hidden. Meta said Muse is the first AI agent covered by Link’s purchase protections. Support for Shop Pay and 1Password is also planned.
Meta said users can review an audit trail of Muse’s activities, disconnect services or change permissions at any time. They can also instruct the agent to forget specific information and opt out of having their interactions used to train Meta’s AI models. The company said Muse conversations and data stored in the virtual machine are not shared with its advertising systems.
Later this year, Meta plans to introduce Muse Confidential VM, which will encrypt the entire virtual machine using a key held only by the user. According to Meta, this will prevent even the company from accessing the encrypted data and conversations.
Muse is initially rolling out in the United States on iOS, Android and the Muse website, with support for AI glasses planned. Meta said the service will be free for most users, with subscription plans available for those who require additional capabilities.
