The proposition
Meta’s App Store description presents Muse as an assistant for email, reservations and shopping. It also describes approval controls, an activity log and revocable connectors. Those are vendor claims, not results from our own testing.
Where the value might be
Our assessment: an agent is most interesting where a task crosses several tools. Preparing an outing could involve a calendar, research and a reservation. A useful evaluation would measure completed tasks, corrections and how clearly the assistant explains an action before approval.
The sensible first trial is a bounded research task. Check its answer against the original sources before delegating a consequential action. We have not measured task reliability, time savings or value for money.
What to watch
The app listing warns that Muse can make mistakes or behave unexpectedly. Connecting more accounts broadens what it can do and what information it can access. Our view is that permission clarity deserves as much attention as the quality of the answer.
The model is a separate story
Meta announced Muse Wonder on October 12. Its research post says the models are available through developer products, with Muse app experiences to follow, beginning with subscribers. An announcement does not establish which model a particular app session uses.
Our provisional verdict
Muse has a concrete proposition: move from advice to completed tasks. A recommendation needs hands-on evidence about reliability and approval controls. This is a source-based assessment, not a hands-on review. App prices, regional availability and referral rewards are not verified here.