Meta launches Muse, a personal agent with browser access and approval gates
Meta's new Muse app can browse, fill forms, make purchases, and keep working in the background, with user approvals and an audit trail for consequential actions.
Meta introduced Muse on September 8 as a personal AI agent powered by Muse Spark 1.3. The Android and iOS app is designed to carry out multi-step tasks across the web and connected services, rather than stop at a conversational answer. Meta says Muse can keep working after the app closes, while the launch page makes access free up to a usage limit. 12
What launched
| Signal | Confirmed detail | Action window |
|---|---|---|
| Browser-based execution | Muse runs in a "Muse Secure VM" with its own browser. Meta says it can book appointments, fill out forms, handle customer service, browse the web, make purchases, create documents, and set reminders. 2 | Test one bounded task from request to completion, including a failed page load or an ambiguous form field. |
| Approval before consequential actions | Muse asks for permission before sending messages, making purchases, or sharing information. The controls include allow once, always allow, and deny, and the product keeps an audit trail of what the agent did and plans to do. 2 | Check which actions pause for approval, how clearly the request states the consequence, and whether the audit trail is enough to reconstruct a mistake. |
| Credential and payment separation | Meta says Muse stores logins in a secure credential store that the agent cannot read. At checkout, Muse can create a one-time card number so the agent and merchant do not see the user's real card number. 1Password integration is listed as coming soon. 2 | Use a test account and a low-value purchase flow first. Treat the pending 1Password integration as unavailable until Meta ships it. |
| Connected apps and background work | Muse can connect to email, calendar, Instagram, and other apps. Meta says Muse can track goals, suggest ideas, build tools for missing tasks, and continue background work after the user leaves the app. 2 | Review every connected account and observe what data the task needs before enabling background execution. |
| Consumer access with a usage ceiling | The app is available on Android through Google Play and on iOS through the App Store. The free tier has a usage limit; users must wait for the limit to refresh or upgrade to a paid subscription after reaching it. 2 | Confirm regional availability, the reset behavior, and the paid plan's actual limits before treating Muse as a daily operator. |
Why it matters
Muse puts the useful question one step past "Can the model answer?": can the agent complete a task across websites and services while keeping the user informed at the points that matter? The product's browser sandbox, approval controls, audit trail, and payment isolation are attempts to make that workflow usable for ordinary tasks. Those controls also make the first evaluation concrete. Measure completion quality, interruption frequency, approval clarity, data access, and recovery after failure. The launch establishes a hosted consumer agent with explicit guardrails; it does not establish that Muse can safely handle every connected account or unattended purchase.
References
- 1
- 2Meta: Muse — Meta's personal AI agent
ai.meta.com
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.
Related content
More from this channel
- OpenAI launches Data agent in ChatGPT Work for natural-language enterprise analytics
- OpenAI releases GPT-Live-1 to the API with full-duplex voice and modular delegation
- DeepSeek launches V4.1-Flash with asymmetric prefill, compressed KV cache, and open weights
- GPT-6 Astra starts rolling out with gated cyber capabilities
- Meta's Muse Spark 1.3 cuts reported tool use and token use for coding agents
- Gemini 3.8 Flash arrives at 3.7's introductory price, with Cyber access restricted
- Claude Fable 5.1 goes public while Mythos 5.1 stays behind trusted access