ChatGPT Ships Sites in Codex, Enabling Full-Stack Apps with Persistent Storage.

ChatGPT has launched Sites, a new capability within Codex that allows users to build and host full-stack web applications with persistent storage and authentication. The feature moves beyond static page generation to enable true application workflows.

Corey Ching, an OpenAI product leader, announced the rollout on LinkedIn this week, emphasizing that the functionality supports real backend infrastructure. His walkthrough demonstrated several practical examples: a forecasting dashboard, an account hub, an event site, and a decision workflow tool.

Sites represents a significant expansion of Codex's capabilities. Rather than generating code that users must deploy elsewhere, Codex can now handle the entire stack, from frontend interface to database persistence and user authentication.

This positions ChatGPT's Codex agent as a direct competitor to development platforms like Replit, Lovable, and Bolt. The key difference: ChatGPT also handles hosting, removing the friction of finding a deployment target.

Ching's examples show breadth. A forecasting dashboard implies data visualization and calculation layers. An account hub suggests multi-user state management. A decision workflow points to conditional logic and persistent records. Each requires backend infrastructure that Codex apparently now provisions automatically.

Sites is currently in preview for ChatGPT Business and Enterprise customers. Ching noted that additional plans will gain access later, though no timeline was specified. This staggered rollout mirrors OpenAI's approach with other Codex features, which recently expanded to mobile for remote development workflows.

The Business and Enterprise focus suggests OpenAI sees immediate value in team and organizational use cases. Internal tools, client portals, and workflow automation are natural fits.

The addition of persistent storage and authentication indicates Codex now integrates with backend services, either through OpenAI's infrastructure or partner integrations. Authentication alone requires session management, token handling, and security protocols that go well beyond code generation.

For small business owners and no-code builders, this removes several technical barriers. Setting up a database, configuring authentication providers, and managing deployments have historically required either technical expertise or stitching together services like Supabase, Vercel, and identity providers.

If Codex handles this stack automatically, it lowers the floor for building functional web applications. The tradeoff: users may have less control over infrastructure choices, vendor lock-in, and cost structures compared to composing their own stack.

The timing is notable. As Codex rate limits recently shifted to a banked reset model, OpenAI appears to be simultaneously expanding what users can build within those limits. Sites could drive higher usage among paying customers by making Codex outputs immediately deployable rather than requiring additional setup work.

Ching's walkthrough video will likely provide more technical detail on how Sites handles database schemas, authentication flows, and deployment configurations. For now, the core signal is clear: ChatGPT is moving from code assistant to application platform.