Meta Muse: A Personal Agent, Explained
Background on Meta’s September agent launch, how it differs from Muse models, and what Meta has announced about access and expansion.
- Announcement
- Sep 8, 2026
- Reported here
- Oct 26, 2026
Muse is Meta’s personal agent
Meta introduced Muse on September 8, 2026. The personal agent plans and carries out tasks across apps a user connects, using a dedicated Muse Secure VM and browser. Meta says it can keep working after the app closes and return for decisions before sensitive steps. This is background on that launch, not a new October release.
The agent and its model have different names
Muse is the agent product; Muse Spark is the model Meta named at launch. Muse Wonder and Muse Wonder Full, announced October 12, are separate models for agentic tasks. Meta said Wonder would later power more Muse experiences. These are distinct things to evaluate, not interchangeable names.
Availability and announced expansion
At launch, Meta described a US rollout on iOS, Android and muse.ai, with free core use and subscription options. Check the official product for current eligibility and pricing. In its September 24 Connect recap, Meta announced additional connectors and planned AI-glasses access in the coming months. That roadmap did not confirm immediate glasses availability.
Scope and user control
Meta says users choose connected apps and permissions, can disconnect services and can request that Muse forget information. For your own evaluation, try a bounded task and check the result, interruptions and time spent correcting it. A useful review should say what you actually asked Muse to do, rather than infer quality from its underlying model.
Go to the source.
Official announcements are the basis for this report. They describe the providers’ claims; they aren’t independent performance tests.