Greg Isenberg compares Meta Muse connectorsAn application connector gives an AI system or workflow a defined way to read data from and perform actions in an external application. with the early mobile app store, arguing that an assistant may bring outside services into a customer's task. He explains the basic pattern with a hypothetical podcast studio booking: the assistant asks a service for availability and priceAgent tool use lets an agent call bounded external capabilities so it can retrieve information or change state beyond generating a response., presents an offer, and acts after the customer approves itAn agent permission boundary defines and enforces which resources and actions an automated agent is allowed to use.. The episode also acknowledges that actual demand and discovery through the connector directory remain uncertain.
Four proposed businesses illustrate different ways to make that connection useful: timely leads for local suppliers, appliance repair dispatch, local padel matches and court reservations, and family dinner planning linked to grocery shopping. Each depends on reliable underlying data or participating providers. The examples are business proposals, not reported operating results.
For finding users, the episode considers relevant creators, sharing built into a service's normal workflow, existing connected marketplaces, and directory placement. Its practical advice is to identify an initial audience the founder can reach directly, since the frequency with which an unknown connector will be recommended has not been established.
For a first version, the suggested process begins with one customer task and a narrow availability or quote flow. The discussion calls for clear prices, authorization, secure credential setup, and tests for booked slots, expired quotes, duplicate requests and booking changes before seeking a public listing. A coding agentAn AI coding agent is a tool-using AI system that can inspect, modify, and validate software within a repository. is presented as a way to build the small service, with human review of its behavior still requiredHuman-in-the-loop AI keeps a person involved in reviewing, correcting, approving, or guiding an AI system's work..
Watch on YouTube




