Meta Muse AI Connectors: The App Store for AI?
Publicado no IA Dots ·
Agentes de IAAI Signalidioma-enInglêsMuseyt-84q4WA3kA8Q
Vídeo em Inglês, publicado pelo canal Greg Isenberg.
Video transcript
Transcript from available captions; it may contain errors.
I think a gigantic opportunity has just opened up and we need to talk about it. Meta has opened up Muse, its personal AI agent to developers, and you can now submit something called a connector, which gives Muse a way to use your service when someone asks for help. And I think this could be the App Store moment for AI. So, if you're thinking about starting a business or you already have one, it's worth understanding what just became possible cuz billions of dollars are going to change hands. By the end of this video, you'll understand what Muse connectors are and where I'd look for business opportunities. We're going to walk through how you can build one using a coding agent and submit it to Meta. We'll also spend some time on how people actually find what you've built. Because let's be real, that's a big part of whether any of this turns into a business. I'll give you four Muse connector startup ideas for free by the end of this video. And if you're new here, my name is Greg Eisenberg and I'm the host of the Startup Ideas podcast and I know a thing or two about social apps. In the past life, I was an advisor to TikTok and advisor to Reddit. And now I build AI native businesses. If this episode gets 7,000 likes, comments, and subscribes, I'll drop a Muse connector guide 100% free with more ideas. I'll include the link in the pinned comment description if that happens. So, go ahead and do that. All right, let's start the episode. >> The Startup Ideas podcast. >> [music] >> It's sippin' time, baby. >> Think about what happened when Apple opened the App Store in 2008. Apple had built the phone. Now an outside developer could build something useful for the people who owned it and sell it through the store. Someone can build a photo editing app and charge for it. Apple supplied a way to distribute the software and collect payment so that the developer could focus on making something people wanted to use. And it was a win-win scenario, by the way. Apple got a way better App Store and developers made money. By June 2010, less than 2 years after the store had opened, Apple had paid developers a billion dollars from paid apps and in-app purchases. So, this new market had formed around the phone and outside businesses were earning a bunch of money from it. That's why I'm interested in this Muse announcement. Meta is building the assistant and other businesses could supply services that help it complete a customer's request. How big that market will become will depend on whether people actually use Muse for those jobs. But the opening for developers is now. And yes, I know someone in the comment comment section is saying, "Well, who knows if this is going to be a big app?" No one really knows, but it is today, as of recording, the number one app in the App Store. When I'm thinking about the App Store analogy versus the App Store for AI analogy, there's another interesting wrinkle. So, your service can become useful halfway through a bigger request. Let me explain what I mean by that. Take someone who's organizing a small event. They're going to need a venue and later they realize they need maybe some equipment that they need to rent. A company that handles equipment rentals could become relevant at that moment. Even though renting equipment wasn't what started the conversation. As a founder, as someone who's trying to build something, I would closely look at those moments. Someone is trying to accomplish something and there's a step where they need access to a business that can actually deliver and money's changing hands. Meta's platform now has a submission process and a directory of approved connectors. There's a bunch of big companies in there now and I think there's going to be a bunch of small and medium-size soon. Editors also can feature them. That gives something real to explore. I remember when I was launching apps in the early days and I got featured, and I was getting like 40,000 downloads every single day from that feature. It was all free. With Muse AI connectors, what we still need to learn is how people will find those services and how often they use them, but I wouldn't be surprised if the directory generated a bunch of customers. That's why being early could help you learn faster than other people in your category. So, I would use that time to understand a customer's request and make my service work really well for it. And then, who knows? You might get featured. Um but we're also uh towards the end of the episode we're going to talk about a bunch of other techniques to get uh downloads or not really downloads, to get usage uh to your uh connector that isn't reliant on Meta featuring you. So, how does a Muse connector actually work? Say I tell Muse I need a podcast studio on a Tuesday afternoon for 2 hours, and my budget is $200. It needs to find out which city I'm in and what kind of recording I'm going to be doing. Then, it needs a business that can answer whether a suitable room is available. An API is a way for one piece of software to ask another piece of software for something. It's just at the core of it. You can think of it as a service window at the studio. You make a specific request, like checking Tuesday's availability, and you get the specific answer back. The connector lets Muse use that window. So, the studio system might return an available room at, I don't know, $160 including the equipment I need. Muse can show me the offer and once I approve it, the service can make the booking and then it sends back a confirmation. That's a proposed workflow, but it shows what the connection is there to accomplish. The studio still has to keep its uh calendar accurate, right? Someone has to make sure the equipment works, of course. If a customer needs help getting into the building, the business needs a way to handle that, too. What's cool is an agent makes it easier to ask for the service. The service still has to be good, though. You can see a real version of this in travel. So, there's a company called Duffel and they say their Muse integration lets people search flights and manage bookings. Duffel also said that its business has passed a billion dollars in annual transaction value on their, I guess, web app and mobile app. So, now they're betting on the Muse AI connector thing. That's the value of bookings, by the way, passing through its business rather than Duffel's revenue and it covers the whole company. But, it's a useful example of a company that already provides something valuable becoming accessible through a new interface. If you built software before, I'd start by looking at what your customers can already do through your API. There might be a useful connector sitting inside the product that you already have. So, if I wanted to start a Muse connector business, where would I actually start a business? Here are four ideas I want to walk you through that I think are pretty good. The first idea is just doing lead gen for business suppliers. Let me explain what I mean by that. Say you run a linen service in Miami. You supply restaurants with clean table cloths and napkins. You could ask Muse which restaurants are opening nearby that might need our service. And your connector reaches a service you've built to track those openings. It brings It brings back relevant restaurants with a source showing when each is expected to open and a public business contact when it's available, of course. The customer can see why each restaurant might be worth approaching. I'd start with one city and maybe just one supplier. You can use sources you're allowed to access, by the way, like, you know, local business announcements, and then you verify the details before adding a restaurant. A planned opening might get delayed. Someone might announce a second location when you're still showing the first. So, keeping that information current is going to be a big part of what you're selling here. General contact search already exists. I mean, I'm sure you've seen things like Apollo. But, your opportunity is knowing something timely about a specific market. So, a restaurant preparing to open gives the linen company like a concrete reason to start a conversation. Although, you still have to find out whether it needs a supplier, of course. I'd show a linen company a few verified opportunities and I'd ask whether it would pay for fresh ones every week, every 2 weeks, every 3 weeks. And then you can charge a subscription. Just a hypothetical example here, but like 100 customers paying $99 a month would give you $9,900 in MRR before any of your costs. So, you can create a nice little business here doing lead gen for business suppliers, and I chose local just as an example because I think it's niche. So, that's startup idea number one. Startup idea number two is home repair dispatch. So, they tell Muse, I don't know, "My dishwasher's broken. Find someone who repairs this exact, you know, Samsung model and can come tomorrow. So, your service is going to go check with, you know, participating repair companies and it returns the ones that handle that specific appliance. So, Samsung, this, you know, model and that obviously they have confirmed availability. They can actually go and fix the thing. The customer then chooses a provider and agrees to share the job details and your service passes the request along or it could book the appointment where that's supported. So, you could charge the repair company like an agreed-upon fee, $100 let's say, for a qualified intro or a confirmed booking. Thumbtack, big company, they already validated that people will pay for customer leads. So, your service would have to deliver requests those providers actually want. I'd begin with, you know, one type of repair, one city, one town. You get to know the providers and what they can actually handle. A confirmed opening tomorrow is useful information when someone is standing in a kitchen with a broken appliance. That homeowner might only need you occasionally. The repair company needs that steady flow though. They need customers every single day. And that's why I'd look at demand across the local market as well as how often one person actually comes back. Startup idea number three is a fun one, you know, if you know me, I like racket sports and, you know, where I live, paddle is getting really, really popular. So, startup idea number three is a local paddle match and court finder. So, someone opens up Muse and they ask, "Hey, find me a paddle game tonight near me with people around my level." So, your service is going to go looking look for open spots at participating clubs and and courts, and it's going to show suitable games and how much it's going to cost. The player is going to approve the reservation. If they already have a group, they could ask for a court instead and share the booking details with their friends. I'd imagine, by the way, that Muse is going to get more and more social over the next 6 to 12 months. I mean, we're talking about meta, right? They understand social better than anyone. I would start with one city. You know, I think it's called Playtomic. They offer court and match booking. So, I'd look for a useful gap, such as local clubs whose availability players currently have to check separately. You need those clubs, by the way, to to obviously let you access their opening and make reservation make reservation. Some some do and some won't. But, you can you know, earn an agreed fee on these bookings. You know, a club with an empty slot has something they they want to sell, right? And someone who plays several times a week has a reason to keep using a service that helps them get a game. Startup idea number four is a family dinner planning and grocery integration connector. So, a parent might say, "Hey, sort out dinners for the next three nights. 20 minutes of cooking each night max. We already have rice and broccoli." So, your service is going to build the plan from recipes you've tested for that specific kind of family. It adjusts the quantities. It turns the missing ingredients into a shopping list that they can go buy from. Instacart, their developer tools actually let you create a shoppable list that the customer opens to review products and finish checkout. So, that gives you like a practical starting point for helping someone get the ingredients home. By the way, side note, maybe someone like Instacart ends up just acquiring this. Like there might be a bunch of of these connectors that get acquired, um you know, over the next couple years if this gets big. If I If I was trying to test this business, you know, I'd test a subscription with families who want to use it every week, right? The meals have to work for them. But if their kids like the dinners and the portions make sense and they have a reason to use your service for the next grocery shop, you know, I'd want to learn what they actually cooked and use that feedback to improve the next plan. You know, of these four ideas, I'd probably investigate the lead generation idea first if I was starting with a small budget. Um you can put together a useful sample and show it to a potential customer before writing much software, and that's what I like about those types of businesses. For For someone who already knows um a bunch of local service providers, the home repair idea might be a closer look, but yeah, I tried to give you, you know, a few different types of startup ideas, more consumer, more more businessy. My guess is the consumer ideas are the most fun to do and because it's meta, they're more of a consumer app company, there's more of those use cases. But uh you know, I think there's going to be a lot of money and also business use cases as well. Of course, this brings us to the question I'd have if I was watching or listening to this, which is great. I have a connector. How are people going to learn about my connector? How are they going to learn about my business? Obviously, getting into the directory, it you know, gets a lot of exposure. Um but whether an unknown service is going to get recommended, you know, during a general conversation is something we still need evidence for. So, I'd build my initial plan around customers I have a way to reach. So, the first idea I have for you is work with existing creators who have audiences. So, for the family dinner service, that could be a creator who shares vegetarian recipes for parents. And then you help them demonstrate how to turn a few meals into a grocery order. They give their audience the tested setup instructions, and you agree on how they'll be paid for customers they introduce. The viewer has seen a task they want to accomplish and understands why they connect the service. I start with one relevant one relevant creator, and I see how many people complete that first task. The size of the audience matters less to me, honestly, than whether the people in it need what you're selling. The second idea I have uh for you is, you know, for the paddle service, for example, one person books a court and shares a page with other three players showing the time and location. So, this this idea is basically product-driven viral sharing. Some people call it product-led growth, too. But basically, it's how do you get one person to share your product? You know, in the paddle service idea, those players get something useful from your product before they ever sign up. So, if you give them an easy way to arrange the next game, you know, you have a possible route to another booking. So, you can choose ideas with this kind of sharing built into the work. You can kind of prioritize based on these multiplayer ideas, because that's going to help you, you know, drive uh a lot of growth. And we saw that, you know, in the mobile era, um you know, a lot of social apps, a lot of apps had, you know, social features to it, and And are the ones that became really popular. Same is true uh in this agentic era. So, you know, what's an example of that? You know, a proposal gets sent to someone, an event invitation gets passed around. You know, make the result useful to the person receiving it, and then some of those people may become customers themselves. That's the hope. So, I'd measure that before I'd call it a growth loop, but it's a practical thing to look for when you're choosing between, you know, two or three promising ideas. The third idea I have for you is distribution through existing connected marketplaces. So, Ticketmaster says eligible events can now appear through its connector without each organizer doing additional integration work. So, an organizer now could reach someone through the Ticketmaster experience while continuing to run their own event business. That makes me curious about where each connected service gets its supply. If you're in if you're in an industry with an existing marketplace, investigate how businesses participate and what information helps customers choose them accurately. So, for some people, that could be the most practical way into this market. You know, for others, building their own connector makes sense because they already have customers or partner who can introduce them. So, I'd make that decision early. You want to be able to point to a specific person or channel that can bring your first customers even while you're hoping the direct directory becomes a meaningful source of demand. Of course, route four is just getting feature placement feature placement in Meta's directory. But, you know, let's be real. You're not going to be able to bank on that. You know, when I had that App Store feature that first App Store feature, you know, it was just like basically somewhat random. I woke up and all of a sudden I was like, "Why is why is the chart vertically?" And I think a lot of people are going to have that experience with the meta directory. You know, you're kind of banking on some product marketing manager in Menlo Park to be like, "This is a good app." So, you're interested in building a connector. How would I approach building the first version? And for me, I would actually start with writing down one thing the customer should be able to accomplish, as basic as that sounds. So, for the paddle idea, I'd start with show me the available courts near me tomorrow evening. And then there's a price, you know, for each one. So, you can take that brief to a coding agent like Claude Code or Codex. And then you just give it the documentation for the system that you're connecting to. And then I would describe the customer request in just like plain ordinary uh English or language. If you're building a new service, start with just clearly labeled sample inventory while you're figuring out the flow. I'd ask it to build the availability check and the quote first. And the result needs to show the full price and how long the price is valid. A booking operation can follow once those parts work, of course. What you're making is such a small service that another piece of software can use. It needs to be hosted somewhere Muse can reach. It also needs a way to check which customer is using it, so people only get access to the information and actions they're entitled to. Otherwise, that would be a bad thing. So, you can ask, you know, Claude Code or Codex or whatever to explain each of those pieces as it builds them, and I recommend that. I'll include a starter brief with the episode so you can adapt to your own idea. By the way, that's insane that you're able to just use Codex or Cloud Code to create one of these, right? In the In the App Store era, if you wanted to create a mobile app, you needed like one or two million dollars and a team of 10 people to create something that was high quality. Now, anyone listening to this watching this can do that. So, that's really cool. And there's two ways to, you know, proceed from here. Um and this part caused some confusion when I was doing some research here. So, Meta documents a custom connector option. So, you can ask your own Muse to connect to a service using its API information. That gives you a route to trying a supported integration before pursuing a public listing. So, you still have to test your own like particular these service, and any credentials should go through the secure setup flow. By the way, Meta has to approve you in order to be a a connector. So, that's also an outstanding question is like how many approvals are they doing a day, a month, a week a week? So, I saw Stack Trees founder published a first-hand account of the form for when you submit to Meta for review. view, and that included options for an API or an existing MCP server. MCP is a standard way to make tools available to an AI application. A lot of you know that if you listen to the Startup Ideas podcast, this channel. Um if you've already built one, that may give you a starting point. The submission also ask asks for information about the product and examples of how someone would actually use it. So, you're going to need documentation and support details. So, you're going to want to check the current form as you prepare it because the program is so new. Before I submitted a connector, I'd make sure the service handles the awkward requests, um or you probably I would imagine get rejected. So, ask for time that's already booked, try an expired code, check that repeated request that doesn't actually create another reservation. And I try the next thing a customer is likely going to ask, such as moving the court booking to another day. You know, that's when you're going to find out how much of the actual job your service can handle. So, you can take a lot of the friction out of building with a coding agent like Cloud Code and Codeax, but you still need to inspect the results, um and and just really understand what the service does before putting customers through it, or unfortunately, I think you'll get rejected. So, if if connectors and Muse is interesting to you and you're and you've and you've gotten here and you're like, "Where do I start?" I If I were starting this week, I'd pick a type of customer I could actually talk to and ask about the tasks they keep having to deal with. I want to hear about the last time it happened. How did they get it done? Where did they have to wait? What did it cost them? That conversation usually gives you a more useful starting point than staring at a blank editor trying to invent an AI business. Or you can go to ideabrowser.com, where there's a bunch of startup ideas, and you can literally it's literally sign up for free, uh and it's a free idea of the day to help get your creative juices flowing. Then I'd choose a small part of the job and see whether I can make it work through a connector. And that would that's what makes this whole Muse launch worth paying attention to. Meta has opened another way for a business to participate in the things people ask their assistant to do. So, we can start learning which services people want there and what they're willing to pay for. And the only like outstanding question is, do we think that Muse is going to have 100 million, 200 million, 50 million, or billion users? That's an outstanding question. And how many of these Muse AI connectors uh are going to get approved in the App Store, right? That's a Is it going to be like YC where they only let in 0.0 1%? Or is it going to be like the Apple App Store? Where most are accepted as long as they are high quality, deliver value to people. I keep thinking about that idea about the club with the empty court tomorrow evening, the paddle court. You know, there's the group that wants to play and a club that would love to fill the spot. And just helping them compete complete that booking is a business problem, you know, you can understand. You know, I'd start I'd start with something like that, you know, for me. Again, I like racket sports. So, to me I understand it. I'd start with something like that and I would see what I can learn. I hope this got the creative juices flowing. That's That's all I could ask for. You know, this is an incredible time to be building, shipping your ideas, shipping your heart out. It's just an incredible time. I'm rooting for you. Until next time, thank you for everything. See you.
Assistir ao vídeo original no YouTube
Assuntos: Agentes de IA, Muse. Publicado no YouTube em 2026-09-24.
Seleção automática por assunto e idioma. O vídeo pertence ao canal de origem.