OpenAI is changing where people discover and use software. At its September 29, 2026 DevDay, the company introduced new plugin extensions, app recommendations, Sign in with ChatGPT, an enterprise marketplace and persistent Dots agents. Taken together, the announcements move software discovery closer to the ChatGPT conversation itself rather than leaving users to search for a separate application first.
The important development is not simply that ChatGPT can connect to more apps. It is that the user can start with a task instead of a product.
A person who wants to revise a presentation, analyze a file or manage a workflow may not need to decide which application to open first. ChatGPT can identify a relevant connected tool during the conversation, while plugin extensions can provide an interface for using that tool without leaving ChatGPT. OpenAI’s developer documentation now describes extensions for sidebar apps, conversation panels, file viewers and editors, settings and deep links.
That creates a new layer in software distribution: the AI interface itself becomes a place where applications can be discovered, connected and used.
How Does OpenAI’s App Model Differ from a Traditional App Store?
The traditional app model generally asks users to make the first decision: find an application, install or open it, sign in and then complete the task.
OpenAI is moving toward a different sequence:
Task → AI identifies capability → app or tool connects → work happens inside the AI interface
The distinction matters for developers because the application may no longer be the first thing a user encounters. The recommendation can come from ChatGPT after the user explains what they need.
OpenAI’s plugin directory already supports browsing and plugins suggested during conversations, while individual plugin availability depends on factors such as plan, workspace, role and region.
For app companies, that potentially changes the acquisition funnel. Instead of relying only on web search, direct traffic or conventional app-store discovery, they can build an experience designed to be surfaced at the moment a user needs a particular capability.
What Do Plugin Extensions Add?
OpenAI’s plugin architecture is becoming more like a software interface than a simple connection layer.
The company’s developer documentation says plugin extensions can place an application in the ChatGPT sidebar, open it beside a conversation, provide settings, and let users open or edit supported files through the application’s own interface.
That means an integration can do more than retrieve information for ChatGPT.
It can provide an interactive workspace.
TechCrunch reported that Figma and Adobe demonstrated this model at DevDay, showing how design work that traditionally happens in standalone applications can instead be carried out through an integrated ChatGPT experience.
This is one of the clearest changes for developers: integration is becoming part of the product experience itself.
Why Does Sign in with ChatGPT Matter to Software Distribution?
Identity is another piece of the model.
OpenAI’s Sign in with ChatGPT lets users create, link or access supported external accounts using their ChatGPT identity. OpenAI says the identity flow itself shares only basic profile information such as a user’s name, email address and profile picture, where available; conversations, memory, files and billing information are not independently shared through sign-in. Additional data or action access requires a separate authorization flow.
OpenAI is also allowing eligible users to use their ChatGPT plan for AI requests in participating third-party applications. For Plus and Pro users, eligible requests can count against included ChatGPT usage, while the third-party application can still maintain its own subscription, infrastructure fees or premium charges.
That is a meaningful distribution change because the AI subscription can travel with the user into another product.
The application still owns its relationship with the customer, but part of the AI infrastructure behind the experience can come from the ChatGPT account.
What Changes When an AI Agent Chooses the Tool?
This is where OpenAI’s Dots become particularly relevant.
OpenAI describes Dots as always-on agents that can continue working between conversations, operate from their own cloud computer, use connected applications and maintain context for ongoing work.
That changes the role of software again.
A user may not need to identify the exact application required for a task. Instead, the agent can work across connected capabilities and bring the result back for review.
OpenAI’s plugin system also supports event subscriptions through MCP Events. A user can tell ChatGPT to monitor an event, such as a new message or document change, and define what should happen when that event occurs. The connected server then sends matching events to ChatGPT, which follows the user’s instructions.
This moves software beyond the traditional “open and use” model toward “connect and let the system act when needed.”
For businesses, that could matter more than simply having another AI-powered application. The value shifts toward how reliably a tool can participate in a larger automated workflow.
Is OpenAI Replacing the Apple or Google App Stores?
Not yet.
That distinction is important because the recent announcements could easily be interpreted as the arrival of a complete alternative app economy. OpenAI has introduced several pieces of that infrastructure, including discovery, interfaces, identity, integrations and enterprise distribution. But it has not announced a conventional consumer billing and revenue-sharing system equivalent to the core economic model of Apple’s App Store or Google Play. TechCrunch reported that third-party subscriptions and charges still remain part of the ecosystem.
OpenAI’s enterprise Marketplace is also structured differently from a conventional self-service store.
According to OpenAI’s current documentation, eligible enterprise customers can apply part of their OpenAI commitment toward qualifying partner products. During the initial rollout, customers contract with the partner, the partner invoices the customer, and OpenAI handles eligibility and reconciliation. The Marketplace initially supports product discovery and expressions of interest rather than self-service checkout.
That makes the current strategy more precise:
OpenAI is building the distribution and interface layers first. The full consumer app-store economics have not arrived.
What Does this Mean for App Developers?
The change gives developers a new question to answer.
Previously, the product itself was often the destination.
Now the developer also has to consider:
Will ChatGPT Know When My Application Is Useful?
That makes several capabilities more important: clear tool descriptions, reliable actions, sensible permissions, useful interactive interfaces and integrations that work well in AI-driven workflows.
OpenAI’s current plugin system allows ChatGPT to automatically use installed plugins when they are relevant to a request, while users can also explicitly invoke supported plugins.
That creates a form of AI-mediated software discovery.
An application does not disappear. Its role changes. It becomes one of the capabilities available to an AI system when the user’s task calls for it.
Conclusion
The broader significance is the potential shift in the unit of software discovery. For decades, the familiar pattern has been:
Website → App → Feature → Task
AI interfaces introduce the possibility of:
Task → Agent → Capability → Software
That reverses the starting point.
The user does not necessarily need to know which product provides the answer. The AI can become the layer that connects the request to the appropriate tool.
OpenAI’s latest announcements therefore represent more than another round of ChatGPT integrations. They show the company building the infrastructure around AI-mediated software distribution: applications that can be discovered inside conversations, interfaces that can run alongside the chat, identity that can move across participating services, events that can trigger workflows and agents that can operate across connected tools.
The question for the software industry is no longer only which app a customer downloads.
It is increasingly which capability an AI system chooses when a customer describes a job.
