Compare the App Store’s launch with OpenAI plugins across discovery, trust, payments, portability, and gatekeeping, and see where the analogy fails today.
At DevDay on September 29, 2026, OpenAI announced plugin extensions that let developers add interfaces such as sidebar apps and conversation panels to ChatGPT.
Those additions build on a broader platform for bringing third-party software into the assistant, with a shared plugin directory serving ChatGPT and Codex. A plugin packages a useful part of a product so someone can use it while working with an assistant. A plugin might, for example, let an existing customer look up an order or update a project without opening your dashboard. Developers provide the capability; the customer can install it and bring it into a conversation.
For an independent developer or a small software company, a shared directory could provide another route to customers. The combination of installable capabilities and a platform review process invites comparison with the App Store. The business question is whether developers can acquire repeat users and earn money under the platform's rules.
Where the analogy holds #
A plugin can contain skills, which package reusable instructions and resources, an MCP server that exposes tools, or both. Public plugins are published once and become discoverable on supported surfaces in ChatGPT and Codex. OpenAI's quickstart documents both the package and the shared directory.
Developers can connect a deployed MCP server, install a personal plugin, and invoke it with @ in ChatGPT Work. Custom UI is optional. A useful plugin can therefore start with one focused capability, such as retrieving an account record, before adding a custom interface.
Dedicated apps and websites remain useful for complex configuration, visual work, and access to audit history. Some workflows may instead begin with a request to an assistant, which then selects an installed capability.
OpenAI's submission process requires publishers to use a verified identity and complete automated checks before review. They decide when to publish after approval. MCP submissions require test cases and a working test account when sign-in is required. The plugin guidelines cover privacy and reliability alongside accurate tool descriptions and developer verification.
Apple's App Review Guidelines describe a curated store with expert review and editorial discovery. Both platforms use review and discovery rules to govern third-party distribution.
The comparison, dimension by dimension #
The App Store and OpenAI’s plugin directory both give developers a shared place to distribute software. At the App Store’s launch, customers downloaded apps directly to the iPhone and iPod touch; OpenAI’s directory makes plugins discoverable across supported ChatGPT and Codex surfaces. Both platforms also govern how that software reaches users. Apple uses expert review, malware scans, and editorial discovery, while OpenAI requires publisher verification and review and may offer additional placement or suggestions to plugins with strong utility and user satisfaction.
The commercial comparison is less convincing. Apple’s store launched with free and paid downloads, with purchases charged to an iTunes account. OpenAI currently allows commerce for physical goods and access to digital entitlements customers already own, but prohibits new digital sales. An existing software business can use a plugin to serve its customers through an assistant, though the directory does not provide another place to sell that software.
Apple also reported more than 800 apps available and more than 10 million downloads in the first weekend. OpenAI’s documentation describes how plugins are published and discovered, but provides no comparable demand figures or conversion economics. The analogy is therefore strongest on distribution and review. Whether the directory can support a similar market for paid software remains unproven.
Where the analogy fails today #
The original App Store paired discovery with a transaction system from the beginning. Apple's launch report described iTunes billing and more than 10 million downloads in the first weekend. OpenAI’s current plugin rules describe a much narrower commercial surface.
Under the current plugin guidelines, plugins may conduct commerce for physical goods. They may let existing paid customers use entitlements they already own. They may not sell digital products or services. Subscription starts, upgrade promotions, and direct links to transactional upgrade pages are also prohibited. The checkout documentation recommends external checkout for eligible physical goods and limits the ChatGPT payment sheet to select marketplace partners in beta. It also describes checkout using payment methods already saved with a merchant, without collecting new payment credentials. None of these options establishes a paid digital-plugin marketplace.
For digital products, the current rules permit serving existing paid customers through a plugin. Builders can use that access to improve an existing workflow and measure whether customers return. The documentation cited here does not establish a future paid digital-plugin marketplace, its revenue share, or its conversion economics. It also does not disclose a ranking formula. Demand also needs evidence. A directory creates supply and a place to browse. Sustainable businesses need repeated user intent and workable retention. They also need support economics that leave room to maintain a customer relationship. A successful tool call demonstrates that the integration works. It does not establish how often customers will use it or whether that usage can support a business.
What builders should do now #
Start with one narrow capability that produces a result users can verify. Measure whether invocations succeed and whether users return. Track latency and permission failures separately. Choose tools around complete user tasks.
Keep identity anchored in your own service. OpenAI's authorization guidance expects authenticated MCP servers to verify token signatures, issuer, audience, validity times, and required scopes on each request. These checks preserve an authenticated account boundary between the user and the service.
Keep the underlying capability callable outside one host where practical. MCP can help separate your server from a particular client; authentication, UI behavior, and distribution still need testing in each host. Treat the OpenAI listing and its discovery position as a channel-specific asset. Maintain your web or native interface for workflows that need depth, while the assistant handles intent capture and focused actions.
Evaluate customer acquisition, repeat usage, and support costs under the current rules. Broad digital monetization and repeatable demand would make the App Store comparison stronger.
The directory and review process establish a route to distributing plugins. The sources reviewed here do not yet demonstrate a comparable market for paid software. A focused capability tested with real users can help builders decide whether the channel merits further investment.