Muse - The World’s First Personal AI Agent Built for Everyone
Muse is Meta's new personal agent, it can use a browser, fill forms, handle longer-running tasks after the app is closed, send emails or make purchases after approval, and retain persistent information about the user. Each Muse runs inside a dedicated cloud VM containing the agent and its data. A separate Sentinel agent, isolated from Muse at the system level, decides whether actions are allowed to reach the internet and can escalate sensitive operations to the human. Credentials are stored separately so Muse can use them without seeing them, while users get an audit trail of what it did and plans to do. Architecture (see Security) and mainly the "Muse Secure VM" is also very interesting and maybe more than the agent itself. Can be one of the inspirations for building similar architecture by your-self.
Architecture pattern = Worker agent → isolated runtime → separate policy/security agent → credential vault → approval layer → audit log