Greg Eisenberg — Muse AI Connectors: The Next App Store Moment?

In one sentence

Meta has opened up Muse, its personal AI agent, to developers, allowing them to build "connectors" that enable the AI to perform specific tasks—like booking a court or ordering groceries—by interacting with external businesses.

Why read this (and who it's for)

This summary is for founders, developers, and entrepreneurs looking for the next frontier of AI integration. It solves the problem of "what to build" in the agentic era by providing a framework for identifying high-value, transactional tasks that can be automated via API. You will be able to identify specific customer pain points, evaluate them for "connector" viability, and understand the distribution strategies required to succeed in a crowded AI directory.

The big idea

Greg Eisenberg argues that the launch of Muse AI connectors is the "App Store moment" for AI. Just as Apple opened the App Store in 2008 to allow third-party developers to build on top of the iPhone's hardware, Meta is now providing a platform where businesses can provide the "services" that an AI agent needs to complete a user's request.

The core logic is that while the AI (Muse) is the interface, the value lies in the actionable services—the businesses that can actually deliver a result, handle a transaction, or provide a specific piece of data. Success in this new market depends on identifying "moments" where a user's request requires a bridge to a real-world business. Because these connectors allow for transactional interactions (e.g., "Is this room available?"), they create a massive opportunity for businesses to capture a piece of a transaction that was previously manual.

Key ideas

The "App Store" Analogy

• What it is — The opportunity for developers to build on top of a platform (Meta's AI) to sell services to a massive, pre-existing user base.
• Why it matters — It shifts the developer's focus from "how do I get people to use my app?" to "how do I make my service useful to the AI's request?"
• In the book — Eisenberg points out that in 2010, less than two years after the App Store opened, Apple had already paid out a billion dollars to developers.

The "Mid-Request" Opportunity

• What it is — A service becoming relevant halfway through a larger, complex request (e.g., someone planning an event realizes they need equipment rentals).
• Why it matters — You don't need to be the starting point of a conversation; you just need to be the necessary step that allows a task to be completed.
• In the book — Eisenberg uses the example of an event organizer who starts by looking for a venue but eventually needs a rental company to provide equipment.

The Transactional Connector

• What it is — A bridge where the AI asks a piece of software for a specific answer (like availability) and the business provides that answer and a booking confirmation.
• Why it matters — It turns the AI from a "chatbot" into an "agent" that can actually execute a booking or a purchase.
• In the book — The example of a studio booking: the AI asks for a room on Tuesday, the system returns a price and availability, and the user approves the booking.

The "Lead Gen" for Suppliers

• What it is — Providing timely, niche data to businesses (e.g., identifying new restaurant openings for a linen supplier).
• Why it matters — It creates a "reason to start a conversation" by providing verified, timely information that general search tools (like Apollo) might miss.
• In the book — A linen company in Miami could use a connector to find restaurants opening nearby that need clean tablecloths.

Product-Led Viral Sharing

• What it is — Designing a product so that the act of completing a task naturally encourages sharing with others.
• Why it matters — In the "agentic era," social features and "multiplayer" interactions (like sharing a booking) are the primary drivers of growth.
• In the book — A user booking a paddle court and then sharing the booking details with three other players to complete the group.

Frameworks & tools

The "Connector" Evaluation Framework

Use this to determine if a business idea is viable for a Muse AI connector:

1. Identify the Task: What is the one specific thing a customer wants to accomplish? (e.g., "Show me available courts tomorrow.")
2. Identify the Data Source: Where does the data live? (e.g., a club's private booking calendar.)
3. Identify the Action: What is the transaction? (e.g., "Confirm the booking.")
4. Verify the Friction: Can the service handle "awkward" requests? (e.g., "What if the time is already booked?")

The 4-Step Development Workflow

1. Define the Brief: Write down one basic customer task in plain English.
2. Use a Coding Agent: Feed the documentation of the system you are connecting to into a tool like Claude Code or Codex.
3. Build the MVP: Ask the agent to build the "availability check" and "quote" first before attempting a full booking operation.
4. Stress Test: Test the service with "awkward" requests (expired quotes, double bookings) before submitting to Meta.

Memorable examples & stories

• Duffel's Travel Integration: Duffel (a travel booking company) is using a Muse connector to allow people to search and manage flights. This shows how a company with a billion dollars in transaction value can leverage a new interface to reach customers.
• The Linen Supplier Example: A local business in Miami could use a connector to find new restaurants opening nearby. This illustrates how "niche" and "timely" data can be a valuable product.
• The Paddle Court Example: A user asking for a game near them with people of their skill level. This highlights the "social" and "multiplayer" potential of Meta's platform.
• The Family Dinner/Grocery Integration: A parent asking for a 20-minute meal plan. This shows how a connector can bridge the gap between a plan (recipe) and an action (Instacart shopping list).

Key quotes

• "Meta has opened up Muse, its personal AI agent to developers... I think this could be the App Store moment for AI."
• "Your service can become useful halfway through a bigger request."
• "I'd start with one city and maybe just one supplier... a planned opening might get delayed. Someone might announce a second location... keeping that information current is going to be a big part of what you're selling here."
• "The size of the audience matters less to me, honestly, than whether the people in it need what you're selling." (Paraphrased)

How to apply it

This week

• Identify a "Moment": Look for a task you or someone you know does regularly that requires a "middle-man" (e.g., booking a court, finding a repair person).
• Audit your current APIs: If you have an existing software product, look at what your customers can already do through your API. Is there a "useful connector" hidden inside?
• Find a "Creative Spark": Visit ideabrowser.com to see a "idea of the day" and see if you can envision a connector for it.

Ongoing

• The "Awkward Request" Test: Before launching any connector, intentionally try to break it with "awkward" inputs (expired dates, overbooked slots) to ensure it won't fail in front of a customer.
• Prioritize "Multiplayer" Ideas: When choosing between two ideas, pick the one that allows for easy sharing (e.g., an invitation or a group booking).

Common pitfalls

• Banking on the Directory: Do not assume Meta's directory will be your primary source of traffic. You must have a plan for distribution (creators, viral sharing, or existing marketplaces).
• Over-Engineering the MVP: Don't try to build a full-featured app. Start with one specific thing the customer should be able to accomplish (e.g., "Show me available courts").
• Ignoring Data Accuracy: In local services (like the linen example), the value is in the current status of the data. If a restaurant opening is delayed and your data isn't updated, the connector loses value.

Connects to

• The App Store (Apple): The historical precedent for platform-based distribution.
• MCP (Model Context Protocol): A standard way to make tools available to AI applications, which may provide a technical starting point for connectors.
• Product-Led Growth (PLG): The strategy of using the product's core utility (sharing a booking) to drive new user acquisition.

If you remember only three things

• The "App Store" moment for AI is about providing services (actions) that an agent can perform.
• Look for "mid-request" opportunities where your service is the necessary bridge to complete a larger task.
• Success depends on distribution (creators, viral loops, and existing marketplaces) rather than just being listed in a directory.

Internalize it

• Reflection questions
  • Which "middle-man" tasks do I perform weekly that could be automated by an agent?
  • If I had to build a connector for a local service I understand (like a sport or a hobby), what is the very first "action" a user would want to take?
  • How would I ensure my data remains "current" enough to be useful for a business?
• Flashcards
  • Q: What is the "App Store moment" for AI?
  • A: Meta opening Muse to developers, allowing them to build "connectors" that enable the AI to perform specific business services.
  • Q: What is a "mid-request" opportunity?
  • A: A service that becomes relevant halfway through a larger request (e.g., needing equipment rentals while planning an event).
  • Q: What is the "awkward request" test?
  • A: Testing a connector with inputs like expired quotes or already-booked times to ensure it handles errors before customers see it.
  • Q: Why is "multiplayer" important in the agentic era?
  • A: Because social features and shared actions (like sharing a booking) are primary drivers of growth on platforms like Meta.
  • Q: How can you build a connector quickly?
  • A: Use a coding agent (like Claude Code) and provide it with the documentation of the system you are connecting to.

Fidelity note

Confidence is high. The summary is strictly based on Greg Eisenberg's transcript. One minor point of uncertainty is the exact "approval" rate of Meta's directory, as Eisenberg notes this is an "outstanding question" and the program is very new.