Meta brings Muse to Mac, letting its AI agent touch your files, mail and calendar
Days after topping the U.S. App Store on mobile, Meta's consumer AI assistant expands to desktop with opt-in permission to act inside native Mac apps -- entering a consumer-agent race where a rival, Instinct, is reportedly in talks to raise at a $10 billion valuation.
Meta expanded its Muse AI assistant to Mac on September 18, days after the app launched on mobile and web and rose to the top of the U.S. App Store charts, TechCrunch reported. On Mac, Muse can interact with a user's files, messages, calendar, notes and mail inside their native applications -- access that's opt-in, with the app requesting approval before it takes a sensitive action rather than acting automatically. "The team is shipping fast," Mark Zuckerberg posted on X about the release.
Entering a fast-moving, well-funded field
TechCrunch's report situates Muse's desktop expansion inside a broader consumer-agent buildout: Instinct, a rival personal-AI-assistant startup led by 23-year-old Noah Shinn, is reportedly in talks to raise $1 billion at a roughly $10 billion valuation -- weeks after raising $250 million at a $2.25 billion valuation -- as it hits compute capacity constraints from more than 100,000 users. Both Muse and Instinct shipped voice calling the same week, underscoring how quickly the category is moving from chat interface to an agent that operates software directly.
Why "acts on your computer" is the detail to watch
An assistant that can touch mail, calendar and local files is a meaningfully bigger trust surface than one that just answers questions -- the opt-in, approve-before-sensitive-action design is doing real work here, and it's the detail worth checking rather than taking on faith. For any organization watching consumer AI agents for signs of what's coming to the workplace, the pattern across this launch and Instinct's funding pace is the same one already playing out in enterprise tooling: permissioned, agentic action is where the product race is now, and where the auditing gap -- can you actually verify what the agent did, not just what it says it did -- is going to matter most.