How to Use Meta Muse: A Student's Setup Guide, From First Chat to MCP Connectors

How to use Meta Muse well comes down to one step most tutorials skip: connecting it to the places your work actually lives. On its own, Muse can chat, browse the web, fill out forms and handle email. It has no idea which classes you take, when your problem set is due or what was in last week's lecture. Connectors fix that, and custom connectors let you add services Meta hasn't listed, including MCP servers.

This guide walks through Muse in the order a student should set it up: sign in, change the settings that matter, add connectors, add a custom MCP connector, then put it to work on coursework without crossing your course's AI policy.

How to get started with Meta Muse

Meta introduced Muse on Sept. 8, 2026, rolling it out in the US on iPhone, Android and the muse.ai website (Meta). Before you download anything, check two rules. Muse's supplemental terms bar anyone under 18, or under a higher minimum age where you live (Muse terms), and Meta's Sept. 29 small-business announcement describes Muse as available in the US and Canada (Meta); Meta Japan added Japan on Oct. 23 (Meta Japan).

Then pick where to use it:

• On your phone: the free Muse app for iPhone or Android. You can also message Muse inside WhatsApp.
• In a browser: sign in at muse.ai on any computer. Meta's download page offers apps for Mac and mobile only (Meta), so on a Windows laptop the browser is the way in.
• On a Mac: the Muse for Mac app can find and organize files you ask it to work with and read apps such as iMessage, Notes, Reminders, email and Calendar. You set each app to Off, Read only, or Read and interact, and macOS asks you for Full Disk Access, Automation and Notifications (Meta Help Center). Since Meta Connect on Sept. 23, the Mac app can also drive any app on your Mac once you allow it (Meta), which makes the approval settings below matter more.

To sign in, use a Meta account. If you don't have one, Meta's getting-started page says you can create one with a Facebook or Instagram login, an email address or a phone number (Meta Help Center). Muse opens on the Chat tab. A few basics from that page are worth knowing on day one:

1. You don't have to wait. Meta says Muse isn't turn-by-turn: you can send a follow-up or a correction without waiting for its answer or repeating the context.
2. Side chats keep separate topics apart, such as one per class, while Muse stays aware of both.
3. Reminders. Tell Muse what to remind you about and when, and it confirms the details. Your list is under the Assistant icon, then Upcoming.
4. Files you upload or ask Muse to make, such as a PDF or a spreadsheet, are saved to the Library tab.
5. The usage limit. Muse is free up to a limit; when you hit it, you can wait for it to reset or pay for a plan. Meta's subscription page lists Power at USD 20 a month with 500 million Muse tokens a week, and Max at USD 100 a month with 3 billion (Meta Help Center).
6. Its mistakes. Meta's own help page warns that Muse may be inaccurate or take unexpected actions, so watch what it does, especially early on.

Change these Muse settings before you connect anything

An agent with access to your accounts is only as careful as its settings. Spend five minutes here first:

• Approvals. By default, Muse asks before important actions, such as sending an email or buying something, and you can change when it asks in Settings. Leave it asking. Muse's terms make you legally responsible for what it does, as if you'd done it yourself, and tell you to approve actions like sending messages, paying or deleting data before they happen, because they can be hard or impossible to reverse.
• Read-only access. Meta's connectors page says many connectors can be limited to retrieving data. For email, the launch post says you can allow reading only, or reading and sending. Start with reading.
• Meta's own apps. Facebook, Instagram and Threads connect automatically if they're in the same Accounts Center as the Meta account you use for Muse. Open Settings, then Connectors, to see what's already switched on.
• Memory. Muse remembers context across conversations, which is useful for "my Tuesday lab" and less useful for a personal detail you mentioned once. You can tell it to forget specific things.
• Training. Meta's launch post says you can opt out of your interactions being used to train Meta's AI models. Decide before you paste in class notes.
• The audit trail. Muse keeps a record of what it has done and what it plans to do. Read it after your first few tasks to see exactly what it touched.

How to add a connector in Meta Muse

A connector gives Muse access to another service. Meta's help page on connectors (Meta Help Center) describes two ways to add one from Meta's list:

1. Ask. Type something like "Connect my Gmail" in a chat, and Muse starts the sign-in.
2. From Settings. Open Settings → Connectors, select Connect next to the service, then Continue, and finish any login the service asks for.

For school, the useful ones are usually email, calendar and a notes or files app. Each connector has its own terms and privacy policy, and you can disconnect any of them later. One catch from that same page: disconnecting stops new data, but information Muse already used may stay in its memory and your chat history.

Meta's list is made of consumer and work apps. What it doesn't cover is your coursework, which lives in your school's systems, behind your school's login.

How to add a custom MCP connector to Meta Muse

This is the step that turns Muse from a general errand runner into something that knows your semester. If a service isn't on Meta's list, ask Muse to create a custom connector. Meta's page says Muse walks you through the setup, which may mean getting API details from the service, and keeps the credentials in its Secure Credentials Store.

That page doesn't name a protocol. A service that publishes a remote MCP server can be added this way, and Lykke's own Muse setup works like that. MCP, the Model Context Protocol, is an open standard for connecting AI apps to outside tools and data (MCP documentation). The explainer on what MCP is covers it in plain English, and the free course section on tool use and infrastructure explains how an AI model calls outside tools.

To add an MCP server as a custom connector:

1. Get the server's details from the service: its MCP address (a web link) and how it signs in, usually an API key.
2. Describe it to Muse in a new chat: the service's name, that it's a remote MCP server, its address, and the name of the credential it needs.
3. Give the key in Muse's credential prompt when it asks, rather than typing it into the conversation.
4. Test it with one simple question the service can answer.

Two limits to plan around. First, Muse's agent runs on its own virtual machine in Meta's cloud. Meta doesn't spell out what that means for connectors, but it follows that a server running only on your laptop is out of Muse's reach, so look for one with a public web address. The guide to Canvas MCP servers compares local and hosted servers if you're choosing one. Second, Meta's help page states plainly that it doesn't review custom connectors or how they use your information. The service you connect is the only check, so prefer one that uses a key you can revoke, offers read-only access, never asks for your school password and records what each call did.

Worked example: connect Lykke to Muse

Lykke holds your Canvas classes, synced read-only by the Lykke Chrome extension on a computer, and publishes them as MCP tools at one web address, so it fits the steps above. To add it:

1. In Lykke, go to Settings → MCP Access. The Muse tab is first. Select Create a key for Muse.
2. Copy the message Lykke writes for Muse and send it in a new Muse chat. It carries the four details from the steps above: the name Lykke, that it's a remote MCP server, its address at api.getlykke.com, and the credential name LYKKE_API_KEY. On this page your new key is already in the message, so the key stays in that chat; don't share the chat, and revoke the key if it ever gets out.
3. If Muse still asks for LYKKE_API_KEY, paste the same key into its secure credential prompt. (The version of the message on Lykke's public setup guide leaves the key out, so Muse always asks this way.)
4. Ask "What's due this week?" to check that it works.

Lykke's guide says Muse keeps the connector for your next chats. The Muse guide for students covers the Canvas sync side, read-only keys and what to do if Muse refuses the key.

What to ask Muse for school

Some useful requests need nothing but Muse itself:

• Reminders: "Remind me Sunday at 8 p.m. to finish the lab report." Check that it landed under Upcoming.
• Files: upload your lecture slides and ask "Make a one-page PDF of the key terms in these slides." The file goes to your Library.
• Email, with a read-only email connector: "Find emails from my professors this week that need a reply, and list what each one asks for."

The rest need your classes connected:

• Planning: "Look at everything due in the next 10 days and suggest an order to start them in." If the list is long, sort it by urgency and importance; the free Eisenhower priority demo shows the method. The guide to making a study schedule turns that order into time blocks.
• Studying: "Quiz me on cellular respiration, one question at a time, and tell me which ones I miss." If your class follows the AP Biology outline, the free AP Biology section on cellular respiration is good material to be quizzed on.
• Review: "Turn the questions I missed into flashcards and save them." Spread the reviews over several days instead of one sitting; the spaced repetition guide explains the research and how far apart to space them.
• Catching up: "What did my professors announce this week?"

Notice what's missing from that list: "write my essay" and "answer this quiz". Muse can do a lot of things that your course may not allow.

How to check what Muse did, and undo it

Agents make mistakes, and Meta says so. Build the habit of checking:

• Read the audit trail in Muse after any task that touched another service.
• Check the other side, too. For Lykke, Settings → MCP Access lists each tool Muse used with the reason it gave.
• Watch for steering. A web page, email or shared file can contain text meant to redirect an AI. If Muse proposes something you never asked for, decline and look at what it was reading.
• Revoke, then forget. To remove a connector, disconnect it under Settings → Connectors, delete the key at the service, and ask Muse to forget what it learned from it.
• Turn Mac apps back off. If Muse for Mac no longer needs your Messages or Notes, set that app to Off in Muse's Settings.
• Guard your keys. A key is a password for one service. The free Missing Semester section on security and cryptography covers password managers and two-factor authentication, habits that carry over to API keys.

Before you point Muse at a graded class

Every course sets its own AI rules, and those rules cover Muse just as they cover ChatGPT or Claude. Read the AI statement in each syllabus before you connect an agent to that class. The guide to whether using AI is cheating explains the four common kinds of policy and what to do when a syllabus says nothing.

An agent adds a risk a chatbot doesn't: it acts. Muse fills out forms in its own browser, and on a Mac it can work your apps. Never send it to a quiz, a submission page or a discussion board, and read each approval request before you say yes. A safe default: let Muse handle the logistics and the practice, and do the graded thinking yourself. If an assignment allows AI help, keep the chat so you can disclose or cite it when asked.