What Is Meta Muse? Meta’s Personal AI Agent, Explained

Meta Muse is the personal AI agent Meta launched on September 8, 2026. It does tasks rather than only answering questions: it sends email, books travel, fills out forms and shops for you, and it keeps working after you close the app. It is rolling out in the US on iOS, Android and muse.ai, and you can also message it directly in WhatsApp.

This explainer covers what Muse does, how Meta contains it, where it works, what it costs and what it connects to. It ends with what an association or small nonprofit should check when staff use it with work email. It is a snapshot as of early October 2026.

The short version: Meta Muse runs on its own dedicated computer in Meta’s cloud. A separate Sentinel agent on that machine approves anything that goes out to the internet. Muse is free for most uses, with paid plans for more. Meta added a small business version on September 29, but its announcement says nothing about team or admin controls. If staff connect a work inbox, that is a decision your organization should make on purpose.

One naming note. Muse is the agent. Muse Spark is the model family that powers it, which we covered in our Muse Spark 1.1 explainer in July. Meta also uses the Muse name for developer products, such as the coding tool in our Muse Code post. This post is about the consumer agent.

What Meta Muse does

Meta’s launch announcement calls Muse a "secure, private personal AI agent that proactively helps people meet their goals and suggests ideas." In practice, that covers three kinds of work.

Everyday tasks. Meta’s examples include sending an email and booking travel. Muse can open a browser, fill out forms and negotiate on your behalf.

Longer goals. You give Muse a goal, and it builds a plan, coordinates your time and resources, and moves the work forward on its own. Meta’s examples include selling a car for more, lowering a bill and adjusting a training plan.

Suggestions. Muse can turn a recipe reel you saved on Instagram into a grocery list. It can also remember details about people, such as a friend’s dietary needs.

For long tasks, Muse keeps working after you close the app. It comes back when something changes or when it needs your approval. You talk to it the way you message a person, in the Muse app or in WhatsApp.

When it is time to pay, Muse can check out with Link, Stripe’s payment service. Meta says Muse is the first AI agent covered by Link’s purchase protections.

How Meta Muse runs: the Muse Secure VM and the Sentinel agent

Every Muse user gets a dedicated virtual machine, which Meta calls the Muse Secure VM. Meta says it is "contained so no one else’s agent can reach it." The data and credentials for services you connect are stored there.

A second agent, the Sentinel, runs on the same machine but is kept apart from Muse at the system level. Nothing Muse does reaches the internet unless the Sentinel approves it. When needed, the Sentinel asks you for permission.

Meta’s engineering post, How We Built Safety Into Muse, adds detail:

• Sentinel decides. Muse proposes actions, and only the Sentinel can grant permission to use a connected service or send traffic out.
• Credentials stay out of sight. Login tokens for services you connect are stored in your VM. Meta says the agent never sees the real tokens.
• Email is filtered. Muse filters one-time codes, password reset links and login links out of email.
• Prompt injection gets several layers. These include model training, detection classifiers, and your approval for actions that move data out of the VM.

Muse also checks with you before sensitive actions such as sending an email or making a purchase. It shows a full record of what it has done and what it plans to do.

This is the cloud-first pattern we described in our post on where AI agents run. Muse works from that cloud machine and reaches your accounts through the services you connect.

Where Meta Muse is available

As of the September 8 announcement, Muse is rolling out in the US on iOS, Android and the web at muse.ai. Meta’s page also says you can talk to it "directly in WhatsApp."

Glasses come next. At Connect on September 24, Meta said it is bringing Muse to its AI glasses "in the coming months." On glasses, you will be able to say your agent’s name, and Muse can act on what you are looking at.

Meta’s Connect recap also announced three other additions:

• Voice mode. Muse keeps working in the background while you talk to it.
• Its own email address. Muse can use it to get things done, or you can use it to reach Muse.
• Muse Charm. A small pocket device for talking to Muse. Meta says it will share more later this year.

What Meta Muse costs

Meta’s wording is that Muse is "free for most of what people need, with subscription plans for people who want to do more." Meta’s own pages do not list prices.

According to Axios, the paid plans cost $20 and $100 a month. Axios also reports that Meta’s Alexandr Wang expects most people to stay on the free tier. Treat the prices as reported, not confirmed, until Meta publishes them.

Muse does not show ads, according to Axios. Meta’s own pages say Muse does not share your conversations or VM data with Meta’s ad systems.

What Muse connects to

Meta says Muse launched with "dozens of partners" plus the entire Shopify catalog. Meta’s engineering post says the first connectors cover email, calendar, your car, your home, and Meta apps such as Instagram and Facebook. Meta did not publish a full list on its pages.

At Connect, Meta added a long list of shopping, travel and work services.

The September 8 post also said 1Password support is coming, so Muse can use logins you already have. You choose which apps Muse connects to and how much access it gets. For email, you decide whether Muse can only read mail or can also send for you. You can change access or disconnect a service at any time.

The user controls Meta lists

Meta lists several controls for people using Muse. All of these are product settings or commitments on Meta’s pages.

• Training opt-out. You can opt out of your interactions being used to train Meta’s AI models.
• “Forget.” Muse remembers what matters to you, and you can tell it to forget specific things it has learned.
• No ad systems. Muse does not share your conversations or VM data with Meta’s ad systems.
• No view of passwords or payment methods. Credentials you share go into secure storage, and Muse uses them without seeing them.
• Confidential VM, later this year. Meta plans a Muse Confidential VM, encrypted with a key only you hold, so that Meta cannot access it. Meta says a small group of testers already uses it.

Until it ships, your Muse VM runs on Meta’s infrastructure without that extra layer.

Muse for Small Business

On September 29, Meta announced Muse for Small Business. It is a set of new skills and connectors inside Muse, not a separate app. Meta’s post says it is available in the US and Canada.

The new connectors include Asana, Box, Canva, Dropbox, Figma, Granola, HighLevel, Intuit QuickBooks, Klaviyo, Lovable, Notion, Shopify, Slack, Stripe and Zoom. Muse can also connect to Facebook Pages, Instagram professional account analytics and Meta ad accounts.

Meta’s examples include building a growth plan from sales and social data, reviewing monthly finances, flagging unusual expenses and drafting email replies. Pricing follows the consumer wording: free for most needs, with subscription plans for more. Meta says nothing publishes, sends or spends without your approval.

The post does not mention nonprofits, and it describes no team accounts, shared workspaces or admin controls. Each Muse still belongs to one person.

How Meta Muse compares with OpenAI dots

OpenAI launched dots on September 29, three weeks after Muse. We compared the two in our OpenAI dots explainer. Both run in the vendor’s cloud and keep working between chats.

The main difference is the buyer. Muse is a free-first consumer app. Dots come with paid ChatGPT plans, and business workspaces get admin controls from day one.

What Meta Muse means for associations and small nonprofits

Muse is free, it lives in WhatsApp, and it can read email. Some staff will likely try it before anyone in IT has looked at it.

For a small nonprofit or association, the question is which accounts staff connect. A personal Gmail is their own business. A work inbox in Microsoft 365 or Google Workspace holds member records, donor details and board mail. When staff connect it to a personal Muse, a consumer agent your organization does not manage can read that mail, and possibly send from it.

Muse for Small Business makes this more likely. Its connectors include Slack, QuickBooks, Box, Dropbox and Meta ad accounts, which many small offices already use. As of early October, Meta describes no way for an organization to see or manage those connections.

What to check this month

1. Check user consent in Microsoft 365. Microsoft Entra lets you limit or turn off user consent to apps, and to route requests to an admin for approval.
2. Check third-party app access in Google Workspace. Admins can mark apps as trusted, limited or blocked, and set what happens to apps nobody has reviewed. The default allows users to access any third-party app.
3. Review apps already connected. Look for Muse or Meta entries in the list of apps that have access to staff accounts.
4. Decide on read versus send. If you allow Muse on work mail, decide whether read-only access is the limit.
5. Settle Meta business assets. Decide who may connect your Facebook Page, Instagram account or ad account to a personal agent.
6. Write it down. Add consumer agents to your acceptable use guidance so staff know what is allowed.

Frequently Asked Questions