Meta has introduced Muse, a personal AI agent designed for everyday use. The company intends Muse to help users work toward personal tasks and goals.
How Muse works
Muse runs on Muse Spark, Meta’s model built for real-world agentic work. Users can interact with Muse through the app or directly through WhatsApp, using a conversational interface similar to messaging another person.
Users can tell Muse what they want to accomplish, and it can then develop a personalized plan, coordinate time and resources, and continue the work on its own. It is designed for people without technical experience and can handle everyday tasks as well as larger, multi-step goals.
Muse can:
- Send emails and book travel
- Open a browser and fill out forms
- Negotiate on a user’s behalf
- Continue working after the app is closed
For longer-running tasks, Muse can return when something changes or when it needs user approval, such as before sending an email or making a purchase. Examples include trying to sell a car for more, lowering a bill and adjusting a training plan as other parts of life change.
Memory and suggestions
Muse can remember what matters to a user, make suggestions without being prompted and act on details shared previously, including information mentioned only once. It can then use those details when helping with tasks or making suggestions.
For example, Muse can turn a recipe reel saved on Instagram into a grocery list, suggest a dinner party menu and remember friends’ dietary restrictions before sending invitations. Users can also tell Muse to forget specific information it has learned.
Payments and connected services
Muse can complete purchases using Link by Stripe. Meta said Muse is the first AI agent covered by Link’s purchase protections, which include:
- Coverage for eligible damaged or lost items
- Price-drop protection
- No-fee returns
- A return guarantee for eligible purchases
Link’s wallet for agents generates a one-time-use card, so users can keep their actual card details hidden while Muse makes purchases across the internet. Shop Pay support is coming soon, along with 1Password support so Muse can use users’ existing logins.
Privacy and security
Muse runs inside Muse Secure VM, a dedicated virtual machine that houses the agent and the user’s data. It runs on a dedicated computer in the cloud, while isolation prevents another user’s agent from accessing it and connected-service data and credentials remain securely stored within the VM.
Meanwhile, a separate Sentinel agent runs on the same machine and stays isolated from Muse at the system level. Muse cannot access the internet unless Sentinel approves the request, and Sentinel asks the user for permission when required.
Muse cannot see users’ passwords or payment methods. Instead, securely stored credentials allow Muse to use connected services without seeing the credentials, including passwords that users enter directly into the browser.
Users control which services Muse can access and how much access it receives. They can:
- Choose connected apps and services
- Set the level of access for each service
- Allow email access as read-only or let Muse send messages on their behalf
- Change permissions or disconnect services at any time
- Opt out of having interactions used to train Meta’s AI models
Muse asks for approval before sensitive actions such as sending an email or making a purchase. It also provides a complete audit trail of what it has done and plans to do, while Meta says it does not share users’ conversations or VM data with its advertising systems.
Muse Confidential VM
Meta plans to introduce Muse Confidential VM later this year. The system will encrypt the entire VM, including the user’s data and conversations with Muse, using a key held only by the user. As a result, even Meta will not be able to access the contents of the VM.
Pricing and availability
Muse is rolling out in the US on iOS, Android and muse.ai, with support for AI glasses coming soon. It is free for most needs, while subscription plans will be available for users who want to do more.