OpenAI Turns ChatGPT Into a Place to Find and Use Apps

OpenAI is turning ChatGPT, which the company says now reaches 1.2 billion weekly users, into a place where software can be found, opened and used without leaving the conversation. TechCrunch

Discovery stays inside the chat thread. When ChatGPT recognizes that an app could help with the task at hand, it will suggest one in the flow of conversation. Users can connect that app and start using it right inside ChatGPT.

That model depends on developers building to an expanded plugin architecture, a common framework for adding outside tools to ChatGPT. The setup supports interactive panels where users work with a tool while still talking to ChatGPT. Inputs, saved progress and follow-up questions stay in one place instead of forcing a switch to another tab or separate app.

Identity and paid usage travel with the user. "Sign in with ChatGPT" lets people use their AI allowance inside third-party apps. OpenAI has 16 launch partners for that sign-in method, including Cognition's Devin, Notion, Vercel, T3, OpenClaw, and Dactyl.

The company showed a meeting app running inside ChatGPT that displays upcoming meetings from the user's calendar. Apps can also be shared through lightweight websites made by ChatGPT, where coworkers sign in with their own logins and permissions. For teams, shared output then follows each person's access, not the access of the person who shared it.

Some of this builds on work already underway. Apps in ChatGPT have been available in preview to Business, Enterprise and Edu customers. OpenAI Developers can now submit apps to ChatGPT for review and publication, and users can find apps in the ChatGPT app directory. OpenAI Help documentation also notes a plan to retire custom GPTs across ChatGPT plans and offer migration to plugins.

The broader context here is distribution and control of the software runtime. For two decades, finding software meant a store, an install step and a permission grant from the operating system. OpenAI is testing whether conversational intent can replace that sequence, with the model calling up the right tool at the moment of need and the panel supplying the interface.

Looking at what this means in practice, the questions are practical. How are permissions negotiated when an app is connected mid-conversation. How is data moving between the model, the panel and the outside company's servers logged and audited. How do admins control which apps can be suggested inside Business and Enterprise workspaces, and how does individual sign-in work inside forwarded or embedded lightweight sites.

In my view, the enterprise preview path is the tell. Consumer assistants win on novelty, but platform shifts stick when IT departments can provision, permission and review them. Previewing inside Business, Enterprise and Edu, tying sharing to individual logins and permissions, and routing publication through review and a directory all point toward a model managed by admins rather than open install.

Worth flagging alongside that is what this enables if the plumbing holds. Small internal tools, calendar views, issue trackers, deployment dashboards and document workflows could live where the work is already being discussed, built as panels rather than maintained as separate destinations. For developers, that lowers the cost of reaching 1.2 billion weekly users to building a focused panel and a sign-in integration, not shipping and updating separate apps across platforms.