Apps SDK
The Apps SDK is a collection of tools published by OpenAI that lets developers integrate their own applications directly into ChatGPT. Instead of opening a separate website, users invoke the app within the chat window, where it can also display its own user interface.
A chat program like ChatGPT can, on its own, only generate text. It doesn’t know which apartments are currently vacant, and it can’t buy a concert ticket. For it to do that, it needs to talk to outside services. That is exactly why the maker OpenAI released the Apps SDK in October 2025. The abbreviation SDK stands for Software Development Kit, meaning a toolbox of tools and templates for programmers. Using this toolbox, companies adapt their own services so that they appear right in the middle of a conversation with ChatGPT and can also display images, maps, or buttons there.
The chat as a new marketplace
For companies, this shifts the place where customers show up. Previously, one searched for a hotel via a search engine and landed on a website. If ChatGPT handles the search instead, the user never sees that website. Anyone who isn’t represented as an app in the chat simply doesn’t come up in the conversation.
According to the company, ChatGPT had around 800 million weekly users in 2025. That’s a reach for which the effort of building an app can pay off. At launch, services such as Spotify, Booking.com, Canva, and Zillow were on board. Observers therefore like to compare the move to the launch of the iPhone App Store in 2008.
The comparison falls short in one respect, though. In the App Store, the user deliberately picks the app and taps its icon. In the chat, it’s often the model that decides which app to call up for a given question. Whoever controls that selection holds significant power over competition — this is one of the points of contention that antitrust authorities are watching closely.
From tool call to display in the chat
Technically, the Apps SDK builds on the Model Context Protocol, or MCP for short. This is an open agreement about how a language model communicates with external programs. To this end, the provider runs a server that reports a list of its capabilities, such as “search playlist” or “book a room.” Each capability comes with a description in plain language.
When a user asks a matching question, the model selects a tool based on these descriptions. It sends the necessary information to the server, which responds with data. You can picture it like a menu: the server lists what it offers, and the model places an order.
What’s new about the Apps SDK is the second part. The server may additionally send along a small web interface, which ChatGPT displays directly within the conversation. This way, an actual map appears instead of a list of addresses. For actions involving an account, the user signs in via a standard procedure called OAuth, in which ChatGPT itself never gets to see the password.
What actually reaches the user
In everyday use, you don’t notice the SDK itself. You type “Spotify, make me a playlist for the train ride” and get shown a playable block. Sometimes ChatGPT also suggests an app on its own when it fits the question.
In business news, the term mostly comes up in larger contexts. It’s about the question of whether chatbots are replacing classic search engines as the entry point to the internet. And it’s about money: with the Agentic Commerce Protocol, OpenAI has announced a way to handle purchases directly within the chat, including a share of the revenue.
A common misconception is equating the Apps SDK with the older ChatGPT plugins. The 2023 plugins only returned text and were later discontinued. The Apps SDK, by contrast, allows for custom interfaces and builds on an open standard that competitors also use. Anyone developing for an assistant today can often use the same server multiple times over.