Meta Muse Connectors: The Next App Store Moment for AI?

Meta's message at the top of the video is compressed into one line: the personal assistant Muse is now open to outside developers, and a connector is the bridge that lets an outside service step into a task the user asked Muse to finish. Host Greg Eisenberg frames this as a distribution opening that could redirect billions in flows and builds the discussion for founders, not for model watchers. The note that Muse sits at the top of the App Store charts on the day of recording makes the distribution claim feel less abstract and more immediate.

Why the App Store Analogy Returns Now

The case is made by rewinding to Apple's store opening in 2008. Apple built the phone, outside developers filled it with small utilities like photo editing tools, and the store took care of distribution and collection. By June 2010 payouts to developers had passed one billion dollars; a new economy had formed around the handset. The video argues Muse is assembling a similar assistant layer where outside firms can provide value at the moment of a request; how large that market becomes hinges on whether people actually turn to Muse for those jobs, but the door is open today.

The nuance in the comparison is that a service can turn relevant in the middle of a conversation, not only at its start. Someone planning a small gathering first looks for a venue and only later realizes equipment rental is needed, and a rental firm becomes relevant at that step. The chat did not start with rentals, yet the step where money changes hands requires access to a supplier that can deliver. For a founder, the advice is to watch those intermediate moments closely: intent is clear, the wallet is close, the supplier is one step away.

On the platform side two signals stand out now: a review funnel for submissions and a directory where approved connectors are listed. Big brands are already visible inside, and the host expects small and medium firms to follow soon, with editors choosing features. He recalls getting 40,000 installs a day for free after being featured in the early mobile era and imagines a similar discovery boost from the directory, while admitting no one knows yet how often unknown services will be surfaced. The practical takeaway is to learn faster than others in your niche by being early.

How a Connector Works: From Service Window to Booking

A studio illustration makes the mechanics concrete. The user tells Muse they need a podcast room on Tuesday afternoon for two hours with a 200 dollar budget. Muse first clarifies the city and the recording type, then seeks a business that can answer whether a matching room is free. The application programming interface is described as a service window at the studio: you ask a precise question about Tuesday availability and you get a precise answer back. The studio system might return a room with equipment for about 160 dollars, Muse shows the offer, and after the user approves, the service creates the booking and sends a confirmation. Keeping the calendar accurate, the gear working, and doorway access handled remains the business's job; the agent eases the request, it does not replace service quality.

Travel provides the polished real example. Duffel says its Muse integration lets people search flights and manage bookings, and that the firm has crossed the one billion dollar mark in annual transaction value across its own apps — a figure that reflects booking value flowing through its platform, not the firm's own revenue. The emphasis is on live inventory and real pricing through a direct API link, not scraping. If you have shipped software before, the suggestion is to inventory what your customers can already do through your interface; a useful connector may already sit inside the product you ship.

Four Startup Sketches: One City, One Supplier

The first sketch is a lead radar for business suppliers. Picture a linen service in Miami that outfits restaurants with clean tablecloths and napkins; you ask Muse which restaurants are about to open nearby and your service returns a short list with a source for the expected opening and a public contact when available. General contact search already exists, the edge here is recency in a specific market: a restaurant preparing to open offers a concrete reason to start a conversation. The host advises starting with one city and one supplier, feeding only from announcements you are allowed to use, then verifying before you list — a planned opening can slip, a second location can be mistaken for the first. Keeping that information fresh is a large part of the value. As a smoke test, show a few verified openings and ask whether the buyer would pay for fresh batches weekly or biweekly; the hypothetical math of 100 customers at 99 dollars per month yielding about 9,900 dollars in monthly recurring revenue before costs is used to illustrate how a niche focus can compound.

The second sketch is home repair dispatch. The user reports a broken dishwasher, names the exact Samsung model and asks for someone who can come tomorrow; your service checks participating firms that support that model and confirms availability, the user picks a provider, approves sharing details, and your service forwards the request or creates the appointment where supported. The suggested pricing is a fee per qualified handoff, around 100 dollars, leaning on the precedent that buyers do pay for leads in large marketplaces such as Thumbtack. The starting advice is narrow: one repair type, one town, learn what providers can truly handle. A confirmed opening tomorrow is immediately valuable to someone standing in a kitchen; the homeowner may use you rarely, but the repair firm needs flow every day, so local aggregate demand matters more than repeat usage of a single household.

The third sketch is a local paddle match and court finder, driven by the host's own enthusiasm for racket sports. The user asks for a game tonight near them at a similar level; your service scans open spots at participating clubs and courts and displays suitable games with price, the player approves the reservation, or requests a court for an existing group and shares details with friends. The expectation that Muse will become more social in the next year, given Meta's background, makes the idea more interesting. The suggestion is to start with one city and find a gap, such as clubs whose availability players must currently check one by one; Playtomic already covers many courts, the opening is with those not yet integrated. An empty slot is inventory a club wants to sell, and someone who plays several times a week keeps returning; the permission to access openings and to place bookings is the hinge, with an agreed share on each reservation.

The fourth sketch connects family dinner planning to grocery fulfillment. A parent asks to sort dinners for three nights, 20 minutes of cooking per night at most, noting that rice and broccoli are already at home; your service builds a plan from recipes you have tested for that family profile, adjusts quantities, and turns the missing items into a shopping list that can be handed to a grocery partner. The developer tools of a service like Instacart allow a shoppable list that the shopper reviews and checks out themselves, providing a concrete place to start. The host notes this category could attract acquisition interest. For a test, he would try a weekly subscription with families, watching whether children like the meals and portions make sense, then feeding what was actually cooked back into the next plan. Among the four, the first is flagged as the fastest to validate on a small budget because you can assemble a sample list and show it before writing much code; if you already know local providers, the second feels closer.

Getting Discovered: Do Not Bank on the Directory

The key question is discovery. Listing in the directory brings visibility, but whether an unknown service will be suggested in a general conversation still lacks evidence, so the advice is to build a plan around audiences you can already reach. The first route is working with creators who own attention: a creator who shares vegetarian recipes for parents could demonstrate turning a few meals into a grocery order, providing setup instructions that the audience can copy and agreeing upfront how introductions are compensated. The viewer has seen a task they want to complete and understands why connecting the service helps; audience size matters less than fit, and the metric to watch is how many complete that first task after watching.

The second and third routes combine built-in sharing with distribution through already connected marketplaces. In the paddle example one person books a court and shares a page showing time and place with three other players; the recipients get value before signing up and a path to the next booking appears if arranging the next game is easy — a pattern often called product-led growth that should be measured before calling it a loop. Separately, Ticketmaster is cited as letting eligible events surface through its connector without each organizer integrating separately, raising the supply question of where inventory comes from and which information helps a shopper choose; in some verticals joining an existing marketplace is more practical than building a standalone connector.

Building the First Version and Preparing for Review

The build recipe starts with a single sentence: one thing the customer should be able to accomplish. For paddle, the seed is show available courts near me tomorrow evening with a price for each. Take that brief to a coding agent such as Claude Code or Codex, hand it the documentation for the system you connect to, and describe the user request in plain language. If you are creating a new service, begin with clearly labeled sample inventory while you sort the flow. The advice is to build the availability check and the quote first, ensuring the response shows full price and its validity window, then add the booking operation. What you are shipping is a small service that another program can call; it must be hosted where Muse can reach it and it must verify which customer is calling so people only access what they are entitled to. Ask the coding agent to explain each piece as it builds it.

The submission landscape has two tracks that are often confused. Meta documents a personal connector option where a user can ask their own Muse to connect to a service using its interface details, allowing an experiment before pursuing a public listing; credentials should travel through the secure setup flow. The public track requires Meta approval. The form asks for an interface or an existing MCP server — MCP being a standard way to expose tools to an assistant — plus product information and examples of use, and documentation plus support details. How many submissions are approved per day, week or month remains an open question.

Before applying, the checklist is to stress the service with awkward requests: ask for a slot that is already taken, present an expired quote, repeat the same request to ensure a second reservation is not created, then attempt the natural next action a customer would try, such as moving the court booking to another day. Those attempts reveal how much of the real job the service can handle. A coding agent removes friction in construction, but inspection remains mandatory; skipping it risks rejection. If connectors spark your interest, the closing advice is to pick a customer type you can actually interview and ask about the last time the task occurred, how it was solved, where the wait happened and what it cost. That conversation usually beats staring at an empty editor trying to invent a business from scratch.