# Evolving AI chat with MCP Apps - Phil Nash - NDC Copenhagen 2026

## Executive summary

The talk introduces MCP Apps, a proposed open standard designed to evolve AI chat interfaces beyond plain text. By integrating rich, interactive web UIs (built with HTML/CSS/JavaScript) directly into the conversation flow, MCP Apps allow agents to render mini-applications for tasks like booking hotels or managing playlists. This approach moves interaction from boring 'walls of text' to engaging, visual experiences, making AI more useful for complex user workflows.

## Key takeaways

- The Need for Interactive UIs in Chat: Traditional chat interactions are limited to text (or code/tool calls), which is insufficient for tasks requiring visual exploration, configuration of multiple options, or viewing real-time data. MCP Apps solve this by bringing web-powered interfaces into the chat environment.
- MCP Apps as an Open Standard: MCP Apps is a standard inspired by community efforts (like MCP-UI) and commercial SDKs (e.g., OpenAI's Apps SDK), aiming to provide a unified way for agents to render UIs across different model providers.
- Core Functionality: Sandboxed Web Views: MCP Apps are implemented as sandboxed web applications (HTML, CSS, JavaScript) loaded within an iframe. This isolation keeps the UI safe while allowing it to interact with the agent host via tool calls and a JSON RPC mechanism.

## Technical details

- MCP Protocol Mechanics: An MCP server defines tools, resources (static data/context), prompts, and sampling capabilities. For an MCP App, the definition includes a `resource URI` pointing to the HTML content under the `_meta` property, allowing the host to render the UI.
- UI Interaction Flow: The user triggers an MCP App via a tool call. The tool response instructs the host to load the resource URI into a sandboxed iframe. Within the app, JavaScript connects to the host using a form of JSON RPC to send results (e.g., 'we chose this hotel') back to the agent for context understanding.
- Agent Capabilities and Context: MCP Apps allow agents to perform complex, multi-step interactions (like searching or booking) that are visually guided. The interaction is proxied through the host, ensuring the model understands the user's actions within the visual UI.

## Practical implications

- Build engineers should adopt MCP Apps as the standard for designing complex agent interactions that require visual feedback or multi-step configuration.
- The architecture involves creating a sandboxed web component (HTML/CSS/JS) and defining its interaction points via an MCP server tool definition, allowing it to communicate results back to the host agent.
- Focus on making the UI part of the conversation flow rather than just rendering a link, as this allows the model to understand user intent within the visual context.

## Topics

AI Agents, Conversational UI (CUI), Open Standards, Web Development (HTML/CSS/JS), Tool Calling, MCP Apps, MCP-UI, OpenAI's Apps SDK, Open Rag

Source: https://www.youtube.com/watch?v=CXtiHjCYSH8