Meta Muse Connectors: The New App Store for AI (Full Guide)

Author: Hyperautomation Labs · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Meta Muse Connectors: The New App Store for AI (Full Guide) · Channel: Hyperautomation Labs · 10:28 · Watch on YouTube

Topics: AI agents, Task and workflow automation, Markets and economy

Summary

The video introduces Meta Muse as an AI agent that can use a cloud-based browser and connected apps to complete tasks, while asking permission before actions such as sending emails or spending money. It walks through setup, connector permissions, privacy controls, and a shopping research test, noting both useful capabilities and mistakes. The creator describes potential applications for consumers and businesses, then demonstrates building a custom connector with an MCP server and explains how to submit connectors to Meta’s directory. The video frames the platform as a possible new marketplace, while noting that some announced integrations are not yet live and key details such as fees and review timelines remain unpublished.

Key points

Muse is presented as an AI agent that can browse websites, fill forms, compare options, and continue working after the app is closed.
A separate agent called Sentinel checks actions before they reach the internet; Muse asks permission before sending emails or spending money.
The app includes chat, an activity log, suggested ideas, goals, and a library for generated content.
Users can connect services such as Gmail, Google Calendar, Drive, Sheets, Notion, Spotify, and others, with adjustable permissions.
The video notes that the setting to help improve Meta’s AI models is on by default and can be switched off.
In a standing-desk comparison test, Muse returned a recommendation but its table had errors, so its work should be checked.
The creator highlights possible uses including finding savings and refunds, handling paperwork, planning purchases, and turning work notes into tasks.
The video says Amazon has blocked Muse from its store, several announced retail integrations were not live, and availability was limited to the US and Canada.
A custom connector can be built using an MCP server, added to Muse, and used to retrieve information from a creator’s own data.
Meta’s public connector submission process asks for product details, an MCP server or API, login information, a privacy policy, and terms; fees and review timelines had not been published.

Chapters

Start	Chapter	Summary
00:00	Introducing Muse and its connector platform	The video introduces Meta Muse, its agent capabilities, and Meta’s plan to let businesses and developers build connectors.
01:05	How Muse works and setting it up	The creator explains Muse’s cloud computer, app features, eligibility, plans, and initial setup.
02:08	Connectors, permissions, and a real-world test	The video reviews available and announced connectors, permission and data settings, and a shopping research task that demonstrates both strengths and errors.
04:52	Use cases and the business opportunity	The creator discusses savings, admin, planning, and work tasks, then argues that Muse could create a new marketplace for businesses and connector builders.
07:40	Building and submitting a custom connector	The video demonstrates creating an MCP-based connector, describes Meta’s submission process, and closes with caveats and a guide for viewers.

Quotes

It’s not a chatbot. It’s an agent. — 01:05

So, Muse writes the draft and you decide whether it goes. — 03:13

Nothing added to cart as requested. — 04:20

So, treat it like a smart assistant. — 06:30

MCP is just a standard way to hand a tool to an AI. — 08:13

Concepts and entities

Muse, Sentinel, Meta, MCP, Claude Code, Vercel, Shopify, Duffel, Ticketmaster

Full transcript (original language)

00:00 This is Meta’s new AI agent, Muse, buying something on its own right after it asks for permission. 10 days after launch, it was the number one app in America. And now, Meta is letting any business or developer plug into it. So, in this video, I’ll set it up, connect my apps, show you what it’s actually good at, and build my own connector from scratch because this could be the biggest new storefront since the app store. Here’s how Meta put it on stage.

00:33 People want Muse to easily connect like this to any app or any business, not just big brands. So, last week, we started to open up the connector platform so developers can build connectors on their own. And in less than a week, we received over 1,500 applications with so many amazing names you all know and love. You can see some of them here. Uh I think this could be a generational opportunity to start building for a new platform as it takes off. And the response so far has been really great.

01:05 >> So what is Muse? It’s not a chatbot. It’s an agent. Every user gets their own computer in the cloud. Meta calls it the Muse secure VM and it has a full web browser. You give it a job and it opens websites, fills out forms, compares prices, and keeps working after you close the app. Before it sends an email or spends money, it stops and asks you. And a second agent called the Sentinel

01:36 checks everything before it reaches the internet. The app has a few parts. Chat is where you hand it jobs. The activity log shows every step it took. Ideas are things it suggests on its own, like chasing an airline refund. Goals track bigger things like training for a race. And the library keeps everything it makes from documents to web pages. Setting it up takes about 2 minutes. Right now, it’s for adults in the US and Canada. Get the Muse app on iPhone or

02:08 Android or go to muse.ai on your computer and sign in with your Meta account. the same one you use for Facebook or Instagram. The first thing it asked me was to connect my apps. It’s free with a weekly limit. This is my account on the free plan. If you need more, there’s a $20 plan and a $100 plan. And you can even chat with it inside WhatsApp. Now, connectors. A connector gives Muse access to one of your apps and you decide how much. This

02:42 is the real list in my account today. Gmail, Google calendar, docs, drive and sheets. Notion and granola for work. Open table, Spotify, Pelaton, Philips Hugh and even your Tesla through Tessie. Checkout runs through a wallet with link by Stripe and Shop Pay. Meta also announced Walmart, Best Buy, Sephora, Wayfair, Expedia, Instacart, and GitHub. Those are on the way, but they’re not in

03:13 my list yet. Here’s the part that matters most. Tap any connector and you set the rules. For Gmail, reading is allowed, but sending an email is set to ask. So, Muse writes the draft and you decide whether it goes. Under permissions, you can make it ask before it visits any website. And under data controls, there’s this switch. help improve our AI models. It’s on by default. If you don’t want your chats used for training, turn it off. If this

03:48 is useful, hit subscribe because it’s free. And if you want to back the work, hit join here or the supporter button on Facebook. Now, a real test. I gave it a real job. Find the three best reviewed standing desks under $300. Compare them. Pick one for a small apartment. and don’t buy anything. It opened its own browser in the cloud, went to Best Buy, then to Flexispot, and read the actual product pages. And if you ever want to step in, there’s a

04:20 button to take control of the browser yourself. About 7 minutes later, it came back with a table and a pick, the Bestier 40in desk, $140, which it called the most compact and cheapest of the three. And look at the last line. Nothing added to cart as requested. It isn’t perfect though. The first row is scrambled with the price sitting in the size column and two of the three picks are almost the same desk. So check

04:52 its work like you would with any new assistant. So where is it actually worth using? The biggest wins so far are about money. lower bills, cheaper insurance, refunds people forgot they were owed. Here’s what Meta shared on stage. >> It saved a bunch of people money on car and home insurance. It hunted out refunds they didn’t even know were owed. One person had to negotiate with her cable company through their online customer support chat and it knocked $85

05:23 a month off the bill. Muse has been saving real people real money. Next is paperwork and admin, permission slips, forms, appointments, following up on refunds. It fills the form, then asks before it sends anything. Third is planning and shopping, trips, dinners, gifts, a new stroller. It finds the options, you approve the price, and it checks out with a one-time card, so it

05:57 never sees your real card number. And for work, connect granola and notion, and ask it to turn your meeting notes into tasks or connect Gmail and calendar and ask for a one-page brief every morning. I put 12 prompts like these in a free guide with the connector each one needs. Now, the honest part. Amazon has blocked Muse from its store. Most of the big retail partners are announced, not live. It’s US and Canada only, and it can get things wrong. One reviewer got

06:30 date night ideas for places that had closed years ago. So, treat it like a smart assistant. Let it do the leg work and you approve anything that costs money. Now, why I think this is a massive opportunity. In 2008, Apple opened the App Store and less than two years later, developers had earned a billion dollars from it. Muse has the same pieces. It went to number one on the app store. Meta’s apps reach 3.6 billion people every day. There’s a directory, editors who feature

07:05 connectors, and checkout built-in. And Zuckerberg said the plan is to earn a small fee on transactions. So, Meta wants businesses making sales through Muse. Companies like Duffel for flights and Ticket Master for events are already in. Here’s who wins. One, any business with a booking or inventory system. A salon, a clinic, a restaurant, a plumber. When someone tells their agent, “Book me.” The business with a connector gets the booking. Two online stores.

07:40 Shopify’s whole catalog is already reachable in Muse. So your product details now matter to an AI buyer. Three, freelancers. Most small businesses can’t build a connector themselves, so you can build it for them. And four, anyone with useful data, like my list of free AI courses. Let me show you how easy that is. I gave Claude Code one prompt. Build a remote MCP server with one tool that finds free courses from my list. Test

08:13 it, deploy it to Versel, and give me the link. MCP is just a standard way to hand a tool to an AI. It wrote the server, tested it with a real MCP client, and deployed it. 1 minute and 33 seconds and the link was live. Then in Muse, I pasted that link and asked it to create a custom connector. First, it asked my permission to talk to the new site. I approved it. It tested the tool and saved it for every future chat. Now, I just ask, “Find me free courses on AI

08:48 agents under 5 hours.” And Muse calls my connector and answers from my own data. Custom connectors aren’t reviewed by Meta, so only connect ones you trust. To get into the public directory, go to muse.ai/platform. AI/platform and click submit a connector. You describe your product with example prompts. Give it your MCP server or your API. Choose how people log in and add a privacy policy and terms. Meta reviews it for function, security, and legal. And approved ones

09:23 go into the directory. One honest note, Meta hasn’t published fees or a review timeline yet, and Meta told applicants it got more than 2,000 submissions in the first few days. So, get in early and make sure it works end to end. Real talk, 90% of you watching haven’t subscribed. On Facebook, subscribe is the supporter button. That is what keeps the lights on here. On YouTube, it’s free and it tells the algorithm to send you more. Subscribe to support us. I put

09:55 everything into one free PDF called the Muse Connector Playbook. How to set it up, the settings to change first, 12 prompts with the connector each one needs, what’s live and what’s coming, and the full checklist to build and submit your own connector. Comment connector and my bot sends it to you. New to Claude Code? My beginner guides are linked in the description. And if you want to support this channel directly, hit join under this video or

10:28 the supporter button on Facebook and follow along on Instagram and Facebook for byte-size AI while the deep dive playbooks live here on YouTube. The app store took 2 years to pay out its first billion. This one is just getting started. See you in the next one.