Meta Muse Personal AI Agent
Meta introduced Muse on 8 September 2026, summed up in its own words as it does not just answer questions, it actually does the work: sending email, booking travel, filling forms, opening a browser and negotiating on your behalf, running on a dedicated Muse Secure VM with a Sentinel agent gating internet access, reachable from the Muse app or WhatsApp
Tool Interface
Interactive tool will be available soon
Features
- ✓ It acts: turns long-term goals into action plans and advances the work rather than only advising
- ✓ Muse Secure VM: a dedicated virtual machine housing both the agent and your data, isolated from other users' agents
- ✓ Sentinel approval layer: runs separately and is system-level isolated, so Muse cannot reach the internet without approval
- ✓ It cannot see your passwords or payment methods; shared credentials sit in secure storage and are used, not read
- ✓ Confirms with you before important actions and keeps a full audit history of what it has done and plans to do
How to Use
- Read the official announcement for capabilities and privacy boundaries: https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent
- Talk to it in the Muse app or in WhatsApp, stating goals the way you would message a colleague
- Grant the services you actually want it to reach, and review which accounts and data it asks for
- Keep human confirmation on high-risk actions such as sending email or buying, and revisit the audit history
FAQ
What is Muse?
Muse is the personal AI agent Meta announced on 8 September 2026. Meta's own positioning is that it does not just answer questions, it actually does the work: helping you stay on top of things, taking tasks and projects off your plate, and turning long-term goals into action plans. Official post: https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent
What can it do?
Meta's examples include sending email, booking travel, filling out forms, opening a browser, and negotiating on your behalf within the access you grant. Once you share a goal it helps build a personalised plan, coordinates your time and resources, then advances the work on its own.
How are privacy and permissions designed?
Muse runs on Muse Secure VM, a dedicated virtual machine that houses both the agent and your data and is isolated from other users' agents. A separate Sentinel agent runs on the same machine, system-level isolated, and Muse cannot access the internet unless Sentinel approves. Muse also cannot see your passwords or payment methods; shared credentials live in secure storage and are used without being read.
What if it gets something wrong?
Meta describes two safeguards: Muse confirms with you before important actions such as sending email or buying an item, and you get a complete audit history of the actions it has taken and the ones it plans to take.
Which model powers it?
Muse runs on Muse Spark, which Meta describes as its most capable model so far and one built specifically for this kind of real-world agent work.
Where can I use it, and what does it cost?
The announcement says you can talk to Muse in the Muse app or directly in WhatsApp. Available regions, pricing and age requirements move with the rollout, so check Meta's official announcement and product pages for the latest.