Which Muse is this page about?
This page covers Muse, the personal AI agent Meta introduced on 8 September 2026, which you use in the Muse app on mobile and Mac, at muse.ai or in WhatsApp. Meta's announcement says the agent is powered by Muse Spark, which Meta calls its most capable model to date for agentic work.
Muse Spark is also the name of Meta's model family. Later versions such as Muse Spark 1.1 are offered in the separate Meta AI app and to developers through a Meta Model API preview. Those models, and the Meta AI assistant, are not what this tutorial teaches.
What can the Muse AI agent do at work?
Meta says Muse can answer questions, complete tasks, browse the web, make purchases, generate images, create documents and connect to apps such as email and calendar. It builds an action plan for a goal and keeps working after you close the app, returning when something changes or needs approval.
Muse is designed for personal life first, but Meta also offers Muse for Small Business, a set of skills and connectors for small business owners. For a sole trader, useful jobs include chasing quotes, organising reminders, turning notes into documents and monitoring a supplier's price.
Who can use Meta Muse?
Meta says Muse is rolling out in the US on iOS, Android and muse.ai, and its help centre adds that Muse is in limited testing and not available in all locations yet.
You sign in with a Meta account, which you can create from Facebook, Instagram, an email address or a phone number. Muse is free with a usage limit; paid subscriptions add usage. To subscribe you must be 18 or the age of majority in your country and located where Muse is available.
What is a good 10-minute Meta Muse starter exercise?
Use a fictional freelance photographer, Ines Calder, who sends quotes and forgets to follow them up. Create a short mock list of three invented quotes with dates and amounts.
Minutes 1 to 2: open Muse, tap the menu icon and create a side chat called Quotes so this work stays separate from your main chat. Minutes 3 to 5: upload the mock list and ask Muse for a follow-up plan as a table with client, quote date, follow-up date and a one-line reason. Say you want a table. Minutes 6 to 7: ask it to draft one follow-up email as a document, with a note that it was generated by an AI agent, and tell it not to send anything. Minutes 8 to 9: set a reminder for the first follow-up date and check it under Upcoming. Minute 10: compare the table with your list and ask Muse to fix any mismatch.
How do I get better results from Meta Muse?
Meta's own guidance suggests these habits.
Define boundaries in words
State what Muse may and may not touch, for example: only search my email, not my text messages.
Use side chats per topic
Side chats keep topics apart while Muse stays aware across them, which makes reviewing a single project easier.
Ask for a format and attribution
Request a table, list or file, and ask Muse to add a note that content was generated by your AI agent when it goes to other people.
Tell it to forget
Meta says you can tell Muse to forget specific things it has learned, which helps when a detail is wrong or out of date.
What should I check before letting Muse act for me?
Start with your employer's AI policy. Muse runs on a personal Meta account, so many employers will not allow work email, client files or company systems to be connected to it. Keep the starter exercise to mock data and your own non-confidential material.
Meta says Muse runs on Muse Secure VM, a dedicated cloud computer no other agent can reach, and a separate Sentinel agent must approve anything Muse sends to the internet. Muse asks before sensitive actions such as sending an email or making a purchase, and you can allow once, always allow or deny. You choose each app's level of access, such as reading mail only, and can review a full audit trail.
On data, Meta says Muse cannot see your passwords or payment methods, purchases use a one-time card number through Link by Stripe, conversations and VM data are not shared with Meta's ad systems, and you can opt out of your interactions being used to train Meta's AI models.
Where does Meta Muse fall short?
Meta's own help centre warns that Muse may be inaccurate or take unexpected actions, and asks you to monitor it and intervene. Availability is limited to certain locations while Muse is in testing, and the free tier has a usage limit.
It is a personal agent: the pages we read describe no team administration, so a company cannot centrally set its permissions. Meta says a Muse Confidential VM, encrypted with a key only you hold, will follow later this year.
How does Tosheo.ai teach Meta Muse?
Tosheo.ai teaches personal agents such as Muse in its AI Agents Development category, next to workplace agents, so you learn when a consumer agent fits and when it does not. You practise writing boundaries, reading an audit trail and judging approval requests with no code; the AI agents development course adds optional Code lessons. It is not a bootcamp or a job guarantee.
Lessons take 10 to 15 minutes, use fictional material and end with a task marked against a visible rubric. Start with the 7-day free trial; no credit card is needed.
Sources and scope
Meta Newsroom, Introducing Muse, checked 19 October 2026 ↗
AI at Meta, Muse product page and FAQ, checked 19 October 2026 ↗
Meta Help Center, How to get started with Muse, checked 19 October 2026 ↗
Meta Help Center, About Muse subscriptions, checked 19 October 2026 ↗
AI at Meta, Introducing Muse Spark 1.1, checked 19 October 2026 ↗