OpenAI is building its plugin store on MCP, and the door has a three-week review queue
Max Stoiber, who works on OpenAI's plugin platform, says plugins are MCP servers, and admits the demand: average wait in the developer review queue is three weeks.
OpenAI's application-layer plan came into focus not through a company announcement but through the posts of the engineer building it. Max Stoiber, the well-known open-source developer (creator of styled-components) now at OpenAI, published a call titled "AGI is nothing without its plugins": models need to speak to the systems people already use every day, and OpenAI cannot build every connection itself; the people who know those systems best should build them. The post links to the team's job listing for the Plugin Developer Platform.
The real technical admission is in the replies: asked what plugins actually are, Stoiber wrote that developers will build "interfaces (aka MCP servers) that models can talk to". OpenAI's plugin world, in other words, rises on the Model Context Protocol, the standard Anthropic started and the industry adopted as its common tongue; one of those rare moments when closed rivals plug into the same socket.
The demand side shows in the same thread: to a developer stuck awaiting approval, Stoiber replied that the team is slammed with review requests, that the average wait is currently about three weeks, and that they are automating the review flow and scaling the team to bring it down. The plugin storefront has a queue at the door before it has fully opened.
The bigger picture: this week we covered Nous's agent society and Hermes's MCP command center; now OpenAI is weaving a developer economy around the same standard. While the model race runs on, the real contest is over the socket where models plug into the world.
