OpenAI is opening a new publishing pipeline for apps that run inside ChatGPT, and it is pairing it with a ChatGPT app directory so users can actually find what gets built.
For everyday users, the value is convenience. Instead of tracking down integrations elsewhere, you’ll be able to browse featured apps or search for one by name, then use it in the same chat where you’re already working.
For developers, this is also a directional call. If you were thinking in terms of connectors, OpenAI is steering attention to apps as the main way to extend what ChatGPT can do.
Submissions and review start now
Apps are submitted through the OpenAI Developer Platform for review and publication. The submission flow covers the practical storefront stuff, including directory listing details, testing guidance, and country availability settings. If your app relies on MCP, the process also includes MCP connectivity information.
OpenAI is trying to keep quality consistent across whatever lands in the directory. It is shipping an Apps SDK in beta, plus guidance on what makes a strong ChatGPT app, example apps, an open-source UI library, and a quickstart. The expectation is “chat-native” tools that do one job well, not a messy bundle of features.
Discovery moves into the chat
The directory is designed to live where the work happens. You can access it inside ChatGPT through the tools menu, and it is also available on the web at chatgpt.com/apps.
Once you connect an app, you can run it again by @ mentioning it by name, or by picking it from the tools menu. OpenAI is also testing ways to surface relevant apps inside a conversation using signals like context, app usage patterns, and user preferences, alongside feedback controls.
Monetization is limited, for now
At launch, OpenAI’s guidance points developers toward linking out to their own sites or native apps to complete transactions for physical goods. It also says it is exploring more monetization options over time, including digital goods. If apps add value for you, check out what the best AI chatbots can offer.
The practical takeaway: users should start with one app that replaces a real routine and see if it saves steps. Developers should ship something narrow, submit early, and plan for iteration once the directory rollout begins and real usage shows what sticks.
