--- title: "Meta Muse AI agent: How to use the personal AI agent" description: "Meta Muse is a personal AI agent that can browse, shop, and handle tasks for you. Here's what it does, how to set it up, and what to watch out for." image: "https://images.ctfassets.net/lzny33ho1g45/4zxhoorwmjayK9mkteVsho/e180ad5a4364148c69e711915ada1b02/meta-muse-hero.jpg" --- # Meta Muse AI agent: How to use the personal AI agent Meta Muse is a personal AI agent that can browse, shop, and handle tasks for you. Here's what it does, how to set it up, and what to watch out for. At this point, announcements about new AI agents trigger a very "Florals? For spring? Groundbreaking." response in me—which also means I've gotten very good at tuning them out.  Muse, Meta's new personal AI agent, almost got that response from me too. But it piqued my interest the good ol' fashioned way: with a wide-eyed mascot that triggers a level of cute aggression I haven't experienced since seeing _Despicable Me_. Mascot aside, Muse is actually worth paying attention to. Here's everything you need to know.  **Table of contents:** - [What is Meta Muse?](#what) - [Meta Muse pricing](#pricing) - [How does Muse work?](#how) - [How to use Muse](#use) - [Is Meta Muse safe to use?](#safe) ## What is Meta Muse? Muse is a personal AI agent built by Meta, and it's part of a small but growing category of [AI agents](https://zapier.com/blog/types-of-ai-agents/) that go beyond just answering questions. Muse can actually do things like open a browser, fill out forms, send emails, and negotiate on your behalf. When it needs your sign-off before doing something consequential (like sending an email or making a purchase), it pauses and checks in for human sign-off.  Muse is currently available in the US and Canada on the web, desktop (Mac only), and mobile (iOS and Android), and runs on [Muse Spark](https://zapier.com/blog/meta-ai-muse-spark/), Meta's model built specifically for agentic work.  ## Meta Muse pricing Muse offers three tiers (though Meta hasn't released specifics on what each one unlocks apart from Muse tokens—and there are no details about how those tokens work):  - **Free**: This gives you limited daily/weekly usage. - **Power (from $16/month)**: This gives you 500M weekly Muse tokens. - **Maximum (from $80/month)**: This gives you 3B (yes, billion) weekly Muse tokens. I've seen different prices for the paid tiers, so it's worth checking your subscription options before committing to a paid tier. To do this, open your settings (`command/ctrl` + `,`). In the _General_ tab, click **Upgrade** (under _Usage_).  ## How does Muse work? Muse works similarly to other personal AI agents: you give it a goal, it builds a plan to achieve that goal, and then executes it. For one-off tasks (like booking a restaurant), Muse handles it, and you move on with your life. For something bigger (like running your [side hustle](https://zapier.com/blog/business-to-start-with-10k/)), it keeps working in the background, checks in when something changes, and asks before it does anything you can't take back. Here's what's powering all of this:  - **It connects to your apps**. Muse uses [Connectors](https://www.meta.com/help/artificial-intelligence/1687253048996149/?srsltid=AU7gw4Xj2bDzObH-AJ2Un-sBtFuQeCMeF-tqi4sQLqFRV1ZpoqSTumgT) to connect to third-party apps. As of this writing, it natively integrates with over 40 of them, including every Google Workspace app, Dropbox, and Notion. Muse also connects with Zapier, allowing it to securely access and take action across . You decide what each Connector can do: read-only or full action permissions. And you can disconnect any of them at any time. - **It remembers context**. Like any good AI tool worth its salt, Muse remembers your preferences—like the fact that you're a night owl, not a morning person, so meetings should never happen before noon—and applies them in relevant tasks. - **It checks in before doing anything consequential**. Before Muse sends an email, makes a purchase, or takes any action that's hard to undo, it stops and asks. You also get a running log of everything it's done and everything it's planning, so you can keep tabs on what it's been up to while you weren't watching. - **It can pay on your behalf**. When Muse books that flight or hotel, or pays for anything, it does so using [Link by Stripe](https://stripe.com/payments/link), which keeps your card details hidden from the merchant. In theory, that's what makes it safe for Muse to hit checkout on your behalf, but I would still proceed with caution (for obvious reasons). ## How to use Muse Here's how to get started with Meta's personal AI assistant.  1. Go to [muse.ai](http://muse.ai), and log in. If you already have a Facebook or Instagram account, you can use that. Or you can create an account for free. You can also [download the desktop or mobile app](https://ai.meta.com/muse/download/) and use Muse there. 2. If you're creating a new account, follow the prompts to verify your identity and age. 3. Muse will walk you through a quick setup—like changing your assistant's name and recommending a few apps to connect to. You can also tell Muse which apps to connect to (just chat with it like you normally would). Or you can set it up manually: open the settings window (`command/Ctrl` + `,`), click **Connectors**, and select **Connect** next to your desired apps. You have two main options for connecting: - **Individually connect tools**. This is a good option if you want to connect to only one or two tools since you have to manually connect and authorize each one. - **Securely connect to your entire tech stack via Zapier**. This is one install and gives Muse secure access to 9,000+ apps at once (you decide which ones). It also lets you build repeatable automations directly from the chat window. 4. Follow the instructions to authorize the app. That usually means clicking **Connect **a couple times, logging in to the other app if you're not already logged in, and clicking **Allow** when prompted. Then you'll be redirected back to Muse. 5. Once your connections are set up, tell Muse what you want it to do or give it a goal, and let it take care of the rest (including adding emoji reactions to every message you send). ## Is Meta Muse safe to use?  Meta has put real thought into [Muse's security architecture](https://research.meta.ai/blog/security-and-safety-for-ai-agents-our-approach-with-muse). On the infrastructure side, Muse runs on its own dedicated virtual machine in the cloud, isolated per user, so no one else's agent can reach it. (That being said, we've seen [agents escape their boundaries](https://fortune.com/2026/09/26/openai-ai-agents-secure-sandbox-escape-training-pause-second-time-hugging-face-hack/) before, so take that with a grain of salt.) A separate Sentinel agent runs on the same machine, kept apart from Muse at the system level, and monitors everything Muse does before it reaches the internet. Muse never sees your passwords or payment details—credentials go into secure storage, and Muse uses them without actually accessing them. Your Muse conversations and data don't feed Meta's ad systems, and you can tell Muse to forget specific things at any time. That said, Muse is still new, and any AI agent that can act on your behalf deserves a bit of caution. So it's worth taking a beat before you hand over the reins to a plushy bot. **Related reading**:  - [Meta AI vs. ChatGPT: Which is better?](https://zapier.com/blog/meta-ai-vs-chatgpt) - [Claude for small business: What it is and how to use it](https://zapier.com/blog/claude-for-small-business/) - [Vibe coding security: How to be sure your vibe-coded apps are safe to use](https://zapier.com/blog/vibe-coding-security/)