Meta on Tuesday, September 9, unveiled its new personal AI agent ‘Muse’ that is designed to help users with everyday tasks and projects.

The personalised AI agent is powered by Meta’s large language model (LLM), Muse Spark. It is currently only available for users in the US. The AI agent comes with built-in connectors for a wide range of services.

It is capable of connecting with a user’s email, calendar, payments apps, health and fitness apps, smart home device apps, e-commerce apps, dining apps, and more. With these connections, Muse can carry out various tasks on behalf of the user such as sending emails, booking travel tickets, filling out forms, creating plans, turning recipes into grocery lists, sending party invitations, and making purchases at checkout. Meta said that it plans to add more connectors to the AI agent over time.

On the surface, Muse does not seem all that different from the growing crop of AI agents being rolled out by tech companies. The agentic AI market is already crowded, so Meta’s competitive advantage may have less to do with the agent itself and more with where it could potentially be deployed. The Instagram and WhatsApp parent has the distribution network to make Muse more useful and deeply embedded in people’s digital lives.

However, its biggest hurdle could be Meta’s track record on privacy and data use, which might make some users reluctant to trust the Mark Zuckerberg-led firm with more of their personal information than ever before.

Under the hood

Muse Spark, the AI agent’s underlying LLM, is the first AI model released in April this year by Meta’s costly team (Meta Superintelligence Labs) it assembled to catch up with rivals in the AI race.

Independent evaluations of Muse Spark’s performance showed it catching up with top models from market leaders Google, OpenAI and Anthropic in some areas, like language and visual understanding, but lagging in others like coding and abstract reasoning.

Story continues below this ad

Muse, the agent running on top of Muse Spark, operates in its “dedicated, secure computer with its own browser,” Muse Secure VM, which offers various privacy, safety, and security protections over customers’ data, according to the company.

Meta said that a separate Sentinel agent runs on that same virtual machine, but is kept apart from Muse at the system level. This ensures that Muse does not have access to users’ passwords or payment methods.

Other features of Muse

Muse will continue to work even after the user leaves the app. It has the ability to improve over time by learning from user chats to make suggestions unprompted. However, user conversations and data shared with Muse will not be shared with Meta advertisers.

Users can also customise the Muse agent by giving it a name, picking out its avatar, and configuring its look and various settings that dictate how the agent communicates with them.

Story continues below this ad

Users can choose which apps and services they want to connect to their personalised Muse agent. If a platform does not offer a public API (Application Programming Interface), Muse can set up a connection using credentials the user provides.


When no API is available, Muse can access the service via the browser instead.

Pricing and availability

Muse is being rolled out in WhatsApp, with a Meta AI glasses integration also in the works. The AI agent can also be accessed through standalone iOS and Android apps, and on the web at muse.ai.

The AI agent is free to use but requires users to pay for subscription plans as their usage increases. The paid plans include Muse Power priced at $20 per month and Muse Maximum at $100 per month. In order to alert users about when their free usage of Muse is about to run out, the Muse app features a usage meter that shows users what percentage of their usage they have left.