What if your users never visit your website again - and that's a good thing? As AI assistants become the primary interface for digital tasks, applications need a new way to be discovered and used. Enter the Model Context Protocol.

The Model Context Protocol (MCP and MCP-UI) might be the next revolution in "UI-programming". It might be the connecting link between the AI-hype and frontend programming.

The Model Context Protocol standardizes how AI can interact with external services, data, and tools.
MCP is currently generating significant excitement in the ecosystem of AI-assisted software development (Claude Code, Copilot, Cursor, etc.).

However, this presentation highlights the much greater potential of MCP: in a future world, MCP could make any applications and services accessible to any AI assistant.
The AI assistant could thus take on the role of today's browser: instead of interacting with web servers via the browser, future users could interact with MCP servers via their AI-assistant.

MCP is still in its infancy but is developing very rapidly.
In this presentation, we will use concrete code examples to look at how MCP works and what possibilities already exist today for implementing an interface to an application via MCP. And we will dive into MCP-UI, an emerging protocol to build rich, dynamic interfaces for agentic apps.

This talk is ideal for full-stack developers and engineering leaders interested in preparing their applications for an AI-first future.

Devoxx

A conference talk exploring how Model Context Protocol (MCP) enables AI agents to interact with web applications, going beyond AI-assisted coding. Three integration patterns are demonstrated: embedding AI inside a web app via tool calling, exposing a web app to external AI agents via an MCP server, and MCP Apps which allow AI agents to render custom UI widgets inside chat interfaces like Claude or ChatGPT. The talk also covers Web MCP, a proposed browser DOM API standard (backed by Google and Microsoft) that lets browser-based AI agents interact with web pages through structured tool calls rather than screenshot-and-click automation. Key considerations include discovery challenges (currently app-store-like), new UI paradigms needed, production readiness concerns, and questions about who pays for LLM calls in each model.

Model Context Protocol (MCP): The future of user interactions? by Jonas Bandi