IMAGE: Meta
Meta is giving its AI agent access to apps, services, computers, and eventually the world seen through its glasses. The bigger Muse gets, the bigger the trust boundary becomes.
Meta wants Muse to become an AI agent that can act across your digital life. At Connect 2026, the company showed Muse working across Macs and third-party services, with the agent also coming to Meta’s AI glasses. It can send emails, book travel, make purchases, interact with applications and continue working after a user walks away from the computer.
That ambition creates a fundamental product challenge. The more useful Muse becomes, the more access it needs to the information, services and devices that make up a user’s life.
Meta is building security controls around that access. Muse runs on a dedicated Muse Secure VM, while a separate Sentinel system controls whether the agent can reach the internet. Users choose which services Muse can access, and Meta says Muse asks for permission before sensitive actions such as sending an email or making a purchase.
Meta also plans to introduce Muse Confidential VM later this year. The company says it will encrypt the entire virtual machine, including user data and conversations, with a key held only by the user.
Muse is designed to do more than answer questions. Meta says it can take on tasks, develop plans around longer-term goals, and continue working in the background.
On the Mac, Muse can interact with applications with the user’s permission. Meta is also developing an email address for Muse, allowing users to add the agent to email threads or forward messages for it to handle.
That gives Muse substantially more agency than a conventional chatbot. It can retain information about a user, access connected services, and take actions on the user’s behalf.
The distinction matters because an AI that can act creates different risks from one that simply generates an answer. A wrong response can be ignored. An agent can send the email, book the ticket or make the purchase.
VIDEO: Meta
Meta is building a platform around the agent
Muse’s reach will depend partly on how many external services it can access.
Meta has announced connectors spanning productivity, software development, payments, travel, shopping and other services. The company has also opened its connector platform to developers, creating a path for businesses to build integrations around Muse.
The enterprise potential is particularly significant. Integrations with tools such as GitHub, Notion, Box and Granola point towards Muse becoming another interface for professional workflows. Granola, for example, has been expanding from meeting notes into the broader enterprise AI stack.
Meta is also building relationships across commerce, payments, travel and other transactional services. The company says Muse can use services such as Stripe and Shopify to help users find products and complete purchases.
That makes Muse a potential distribution layer for other companies. If users increasingly ask an agent to find information, manage workflows or complete transactions, businesses will have another reason to build for the platform.
Meta wants Muse on your face
The next interface is Meta’s AI glasses. Meta says Muse is coming to its glasses so users can interact with the agent hands-free and use visual context from the world around them.
The hardware portfolio is expanding quickly. Ray-Ban Meta Audio is Meta’s first audio-only glasses, while the company is also expanding its Ray-Ban Meta, Oakley and Meta Glasses ranges. Meta says it will have more than 100 AI glasses styles across those brands by the end of 2026.
Ray-Ban Meta Audio starts at US$349 and offers up to 12 hours of battery life, according to Meta. Ray-Ban Meta Gen 3 starts at US$449 and adds a higher-resolution camera, longer battery life and new controls.
The significance for Muse is less about any individual frame and more about the interface. An AI agent that lives in glasses can accompany a user throughout the day, with access to both conversational and visual context.
Ming Hua, Meta’s VP of wearables devices, said that sensors on the glasses will disable photo taking or video recording when it detects a user’s attempt to tamper with the recording light on the frame. This was in response to potential invasion of privacy or other nefarious intentions.
Meta’s other big hardware bet is VR
In what is a breakthrough in VR wearables, Meta also unveiled Meta VR Glasses, a roughly 100-gram device designed to combine virtual reality, spatial computing, entertainment and work.
The glasses use a separate module (a puck) for compute, battery and storage. Meta says the device will support a 5K Infinite Display, eye and hand interaction, and a virtual multi-screen workspace. It is scheduled to launch in spring 2027, priced at US$1,299.
Meta is positioning the device as a personal cinema, gaming platform and workspace, extending its hardware strategy beyond AI glasses into a broader computing platform.
The trust boundary keeps getting bigger
Meta’s Connect announcements point to a broader strategy: Muse becomes the software layer, while phones, Macs, glasses and other devices become ways of reaching it.
The company is also making privacy and security part of the architecture. Meta says users can control which services Muse connects to, adjust permissions, disconnect services, and opt out of having their interactions used to train Meta’s AI models.
That becomes increasingly important as Muse moves from answering questions to taking actions, connecting to enterprise software, handling transactions, and operating through camera-equipped devices.
Meta’s biggest challenge with Muse may therefore be the same thing that makes the product compelling: getting users and businesses comfortable giving an AI agent enough access to actually be useful.
