# Sites and the Next Generation of Web Apps | DevDay 2026

## Executive summary

This presentation details 'Sites,' an advanced web hosting and application development platform from OpenAI, designed to allow users to rapidly build full-stack, interactive web applications directly from prompts. Sites goes beyond simple hosting by integrating deep connectivity with user data via plugins, enabling personalized and collaborative experiences. Key advancements include automated data updates using scheduled tasks, enhanced security with built-in authentication (Sign-in with ChatGPT), and support for agentic workflows through the WebMCP standard, allowing AI agents to interact with and update the live application.

## Key takeaways

- Rapid, Full-Stack App Development: Users can generate complex, interactive sites (e.g., launch radar apps) from simple data sources like spreadsheets using prompts, requiring zero coding knowledge.
- Data Freshness and Automation: Sites can be kept automatically updated using scheduled tasks and plugins, which fetch data from external sources (like Google Drive) and update the site's internal SQLite database without requiring redeployment.
- Personalized and Collaborative Experiences: Sites support user-specific data access via plugins; a user visiting the site sees their own data (e.g., their Linear tickets) fetched from their personal connectors, while collaborators can also contribute via editor permissions.
- Agentic Web Interaction (WebMCP): The WebMCP standard allows AI agents to interact with a site via exposed tools, enabling agents to take actions (like searching Slack or updating a gift box) directly within the live application, making the site highly personalized for agents.

## Technical details

- Sites Architecture: Sites are full-stack web apps that come with built-in authentication (Sign-in with ChatGPT) and a SQLite database, eliminating the need to change source code when data changes.
- Plugins and Connectors: Plugins bundle connectors and skills, allowing Sites to access third-party data (e.g., Linear, Slack, Google Drive). This access is user-specific, leveraging each user's existing connectors.
- Scheduled Tasks: Scheduled tasks allow automated data fetching and site updates at set intervals (hourly, daily, weekly) directly from the thread, linking tasks and sites for easier management.
- Custom Annotations & Editability: The new annotation extensibility feature allows developers to embed custom annotations directly into the site's HTML, signaling to Codex how the site can be edited, which is useful for component libraries and design systems.
- WebMCP Standard: WebMCP is a new browser standard enabling agents to interact with a site via exposed tools, allowing agents to query backends directly rather than relying solely on UI clicks.

## Practical implications

- Build internal tools and prototypes (e.g., replacing spreadsheets with dedicated UIs) without traditional development skills.
- Create collaborative sites where different team members can see their own personalized data (e.g., their assigned tickets in Linear).
- Automate data maintenance for internal apps using scheduled tasks linked to external data sources.
- Future-proof web applications by designing them to be easily interacted with and updated by AI agents using WebMCP tools.

## Topics

Web Development, AI Agents, Full-Stack Development, Data Integration, Web Standards, Astra, ChatGPT Sites, WebMCP, OpenAI Image API

Source: https://www.youtube.com/watch?v=QRm7VNMnurA
