Meta launched Muse on September 8, 2026, a personal AI agent available in the United States for users 18 and older through a dedicated app on iOS and Android, via the web at muse.ai, and through WhatsApp. Muse is designed to go beyond chatbot-style conversations and execute multi-step tasks autonomously, connecting to a user’s email, calendar, payment methods, health apps, shopping history, and smart-home services. The product arrives as Meta, OpenAI, Google, Anthropic, and Amazon compete to define what a consumer AI agent actually does and who controls it.
Key Takeaways
- Muse operates across three pricing tiers: a free tier capped at 100 million tokens per week, a Power tier at $20 per month, and a Maximum tier at $100 per month. Meta requires a payment card on file even for the free tier.
- The agent is powered by Muse Spark, which Meta describes as its most capable model for agentic work. Muse Spark shares an underlying model architecture with Muse Spark 1.3, the developer-facing agentic coding model Meta shipped on September 2.
- Muse runs on a dedicated virtual machine called Muse Secure VM, which hosts both the agent and the user’s personal data. A monitoring system called Sentinel controls what actions the agent can take on the internet.
- When making purchases, Muse uses Link by Stripe, which generates a single-use card number so a user’s real payment details are not exposed to merchants.
- Muse carries no advertising. Chief AI Officer Alexandr Wang told Axios that Meta is exploring commerce-related revenue avenues as a future monetization layer beyond subscriptions.
- Meta plans to introduce Muse Confidential VM later in 2026, which will encrypt the entire virtual machine with a key held only by the user, preventing even Meta from accessing its contents.
What Muse Actually Does and How It Differs from a Chatbot
Muse is not a chatbot attached to the Meta AI app. The product is a task-executing agent that connects to a user’s services, plans multi-step workflows, and carries them out with minimal supervision. Meta described the core function in its announcement: once a user shares a goal with Muse, the agent develops a personalized plan and works through it, continuing to operate after the user closes the app. Muse checks back when circumstances change or when it needs explicit approval before taking a sensitive action, such as sending an email, completing a purchase, or lowering a bill.
The range of tasks Meta outlined at launch includes sending and managing emails, booking travel, scheduling appointments, filling out forms (including school permission slips), negotiating bills, monitoring pricing, making purchases, and developing longer-term plans. Meta CEO Mark Zuckerberg framed the product as one that “stays on top of things, takes tasks and projects off their plate, and turns long-term goals into action plans.” Users can name their Muse agent, create an avatar, and customize how it communicates.
Support for Meta’s Ray-Ban smart glasses is planned but not available at launch. No timeline has been disclosed for international expansion beyond the United States.
The Pricing Structure Places Muse Closer to Productivity Software Than Consumer AI
The $20 Power tier and $100 Maximum tier position Muse well above the pricing of typical single-app AI subscriptions. OpenAI’s ChatGPT Plus runs $20 per month, but that product is a conversational assistant, not a task-executing agent with access to a user’s email, calendar, and payment infrastructure. Muse’s $100 Maximum tier is closer to what businesses pay for bundled productivity software, suggesting Meta expects its heaviest users to treat the product more like a personal assistant service than an AI chatbot.
Zuckerberg stated that the free tier covers usage up to 100 million tokens per week, a generous cap for light users but one that could be consumed quickly by an agent running continuous tasks across email, scheduling, and shopping. The paid tiers presumably raise or remove those ceilings, though Meta has not published exact token or task-volume limits in its launch materials.
The pricing also reflects Meta’s broader strategic shift. Meta AI, the company’s free conversational assistant, reached 1 billion monthly active users as of May 2025. Muse represents an attempt to monetize that user base through direct subscriptions rather than advertising, a model the company has historically avoided on its consumer platforms. The fact that Muse carries no ads at launch is a notable departure from Meta’s core revenue engine. For entrepreneurs and business owners evaluating how AI infrastructure is reshaping business operations globally, Muse adds a consumer-facing data point to the conversation about what AI agents cost and what level of access they require.
The Security Architecture and the Trust Question
Muse runs inside a dedicated virtual machine in Meta’s cloud, called Muse Secure VM, which houses both the agent software and the user’s personal data. A separate monitoring system called Sentinel operates on the same virtual machine and controls what actions Muse is permitted to take on the open internet. When the agent makes a purchase, it routes the transaction through Link by Stripe, which generates a single-use card number for each transaction so a user’s actual payment credentials are never shared with the merchant.
Meta has announced that Muse Confidential VM will arrive later in 2026. That product will encrypt the entire virtual machine with a key held only by the user, which Meta says will prevent even Meta from accessing its contents. No specific launch date has been disclosed for the encrypted version.
The security design matters because of the scope of access Muse requires. Granting an AI agent persistent connections to email, calendars, payment methods, health applications, and smart-home systems is a level of data access that goes beyond anything Meta’s social media products have asked of users. Reuters reported before the launch that engineers inside Meta had raised concerns about the potential for the agent to mishandle sensitive personal data. Meta’s response was that Muse runs in a controlled environment and never sees actual passwords or payment details. Those two statements are not contradictory, but they also do not fully resolve the data governance questions that a product of this scope introduces.
The Timing and Its Context
Muse launched 13 days after Meta agreed to an $18 billion settlement with 29 U.S. state attorneys general, who argued that Meta’s social media platforms caused harm to children. The proximity of the settlement to the launch of a product that requests deep access to a user’s personal accounts, payment infrastructure, and daily schedule creates an unavoidable trust context. Meta is asking users to grant an AI agent a level of intimacy with their data that exceeds what any of the company’s previous products have required, and it is doing so at a moment when its record on data stewardship is under active public scrutiny.
The competitive context is equally relevant. OpenAI, Google, and Anthropic are all developing agentic AI products. OpenAI has reportedly been planning tiered pricing for advanced AI agents, with enterprise research agents priced as high as $20,000 per month for professional use cases. Muse’s $100 ceiling positions it firmly as a consumer product, but the product’s ambition, a single agent that handles email, shopping, scheduling, bill negotiation, and long-term planning, overlaps with the territory that enterprise agent providers are also pursuing.
Meta’s distribution advantage is scale. With more than 3 billion users across its family of apps, the company has a path to rapid adoption that no standalone AI startup can match. Whether that adoption converts into subscription revenue or runs into friction around data trust and privacy boundaries is the question Muse will answer over its first year.
Disclaimer: This article is for informational purposes only. It does not constitute financial, investment, or technology purchasing advice. Readers should conduct their own due diligence before subscribing to any AI service or granting third-party applications access to personal data, financial accounts, or payment methods. MarketDaily has no affiliation with Meta Platforms, Inc. or any of its products.
FAQs
How Much Does Meta Muse Cost?
Muse offers a free tier with usage capped at 100 million tokens per week, a Power tier at $20 per month, and a Maximum tier at $100 per month. A payment card is required on file even for the free tier.
What Can Muse Do That a Regular Chatbot Cannot?
Muse connects to a user’s email, calendar, payment methods, health apps, and smart-home services to execute multi-step tasks autonomously. Unlike a chatbot that answers questions in a single session, Muse continues working after the app is closed and checks back when it needs approval for sensitive actions such as purchases or emails.
How Does Muse Handle Payment Security?
Muse routes purchases through Link by Stripe, which generates a single-use card number for each transaction. A user’s real payment credentials are never shared directly with merchants.
Is Muse Available Outside the United States?
No. As of the September 8 launch, Muse is available only in the United States for users 18 and older. Meta has not announced a timeline for international expansion.
