Meta has introduced Muse, a personal AI agent that represents the company’s biggest product bet since the AI glasses: an assistant that doesn’t just answer questions but actually does the work.
An Agent That Does the Work
Unlike chatbots that wait for prompts, Muse is built to take tasks and projects off people’s plates. Tell it a goal and it builds a personalized plan, coordinates time and resources, and advances the work on its own — opening a browser, filling out forms, and negotiating on the user’s behalf. For longer tasks, it keeps working after the app is closed and returns when something changes or when it needs approval.
Examples Meta highlights: selling a car for more, lowering a bill, or adjusting a training plan as the rest of someone’s life shifts. Muse also remembers what matters — turning a recipe reel saved on Instagram into a grocery list, suggesting a dinner-party menu, and recalling friends’ dietary restrictions before sending invites.
It runs on Muse Spark, Meta’s most capable model to date, purpose-built for this kind of real-world agentic work. Conversations work just like messaging another person, in the Muse app or directly in WhatsApp.
A New Kind of Secure Computer
Personal agents that hold credentials and act on the internet need a new security model, so Meta built one:
- Muse Secure VM — each person’s Muse lives on its own dedicated cloud computer, isolated so no other agent can reach it. Credentials go into secure storage Muse can use but never see.
- Sentinel agent — a separate watcher on the same machine, kept apart at the system level. Nothing Muse does reaches the internet unless Sentinel approves, and it escalates to the human when needed.
- Human checkpoints — Muse asks before sensitive actions like sending an email or making a purchase, and shows a complete audit trail of everything done and planned.
- User control — people choose which apps Muse connects to and how much access each gets (read mail vs. send on behalf), can revoke access anytime, and can tell Muse to “forget” specific things it has learned. Conversations are never shared with Meta’s ad systems, and users can opt out of having interactions used for AI training.
For payments, Muse is the first AI agent covered by Link’s purchase protections (built by Stripe): free coverage for damaged or lost items, price drops, no-fee returns, and a return guarantee. Link’s agent wallet generates one-time-use cards so real card details stay hidden. Shop Pay and 1Password support are coming soon.
What’s Next
Meta frames Muse as a first step toward personal superintelligence, which it calls one of the most transformative technologies of a lifetime. Later this year the company will introduce Muse Confidential VM, where the whole VM — data, conversations, everything — is encrypted with a key only the user holds, so not even Meta can access it.
Muse is rolling out in the US on iOS, Android, and muse.ai, with AI glasses support coming. It’s free for most of what people need, with subscriptions for heavier use.