ChatGPT Plugin Extensions Let Developers Build Full Apps in ChatGPT

OpenAI used DevDay 2026 to open up the platform it builds ChatGPT's own features on: plugin extensions, which let outside developers ship what the company describes as entire applications that feel native to ChatGPT and Codex. An extension can give a plugin a home in the sidebar, open interactive panels where people work alongside the conversation, and register viewers for the file types a product supports. Extensions are available on all plans and distributed through OpenAI — putting third-party apps in front of ChatGPT's 1.2 billion weekly users.

This piece explains what an extension can actually do compared to the plugins developers already ship, why the distribution model is the consequential part, and what to weigh before committing a product roadmap to someone else's platform. It is written for developers and product teams deciding whether ChatGPT is now a build target.

Plugin extensions: what OpenAI opened up

The framing OpenAI chose is telling: this is the same platform the company uses internally to build ChatGPT features, now exposed to outside developers. Extensions go well beyond the tool-call plugins of the past two years — a developer can build an editor, a dashboard, or a whole workspace directly into ChatGPT, living in the sidebar rather than appearing only when the model decides to invoke a tool.

The launch demos made the shape concrete: Canva appeared as a sidebar extension, and an Adobe image-generation integration was shown on the keynote stage. Three surfaces define the model — a persistent sidebar home, interactive panels that sit beside the conversation while a user works, and file-type viewers, so documents a product owns open in that product's viewer inside ChatGPT rather than as inert attachments.

How extensions change the plugin economics

Distribution is the part that changes incentives. Plugins previously lived or died on whether the model called them mid-conversation; an extension has a persistent surface a user can open deliberately, which makes it a destination rather than a callback. Combined with the same day's announcements — a conversational Plugin Creator, a streamlined submission process with review tracking, and ChatGPT recommending relevant plugins inside conversations — OpenAI now controls a full acquisition funnel: discovery, install, and daily use, all inside the chat surface.

New MCP Events support closes the loop in the other direction: connected apps can notify ChatGPT when something changes — a new task, message, or content update — and trigger automations, implemented by developers via webhooks. An extension is therefore not just a UI embed; it can be an app that reacts to its own backend state inside ChatGPT.

Why this matters for engineers and product teams

For SaaS products, the calculus resembles the mobile app-store moment: the audience is enormous and the platform owner controls placement, review, and the rules. Sign in with ChatGPT — which went global the same day — reduces onboarding friction to a single identity, and for Plus and Pro users, plan-usage sharing means customers can spend their existing ChatGPT allowance inside 16 partner tools including Devin, Notion, and Vercel. The pieces fit together into a real economy, not just an integration point.

The risk side is equally familiar from app stores: building your primary interface inside a platform you do not control means accepting its review process, its outages, and its strategic shifts. OpenAI shipped its own collaboration suite — Space, Pages, Slides — at the same keynote, so the line between platform features and partner territory will keep moving.

Who should build first

The obvious first movers are products whose file types users already drag into ChatGPT — design files, documents, spreadsheets, diagrams — since a viewer extension turns every one of those uploads into an entry point. Tools with strong notification models (project trackers, support queues, CRMs) benefit from MCP Events, because state changes become automation triggers inside the assistant users already have open.

Teams without engineering budget for a full extension can still climb the ladder: the Plugin Creator builds reusable plugins conversationally from a described workflow, instructions, and reference files, and the improved directory plus in-conversation recommendations give even simple plugins better discovery than they had before DevDay.

Distribution, review and what to watch

The submission pipeline now includes review tracking, requested-fix flows, and updates that do not restart the whole process — table stakes for an app store, new for ChatGPT. What is not yet public is the part that decides business models: revenue sharing, ranking mechanics in the directory, and how aggressively ChatGPT recommends third-party extensions against OpenAI's own features. Those answers determine whether extensions are a channel or a dependency.

Watch the extension docs at developers.openai.com for the technical constraints, how the first wave of sidebar apps monetizes, and whether the in-conversation recommendation engine favors extensions over plain plugins. The platform is open on day one to all plans — the land grab for sidebar real estate starts now.