MCP Apps: Building the Future of AI Interactions

Liad Yosef and Ido Salomon of MCP Apps discuss their standard for building interactive UI components within AI chat interfaces, enabling seamless application experiences across platforms.

Liad Yosef and Ido Salomon presenting MCP Apps at AI Engineer Europe.
Image credit: AI Engineer Europe· AI Engineer

In a significant development for the future of human-AI interaction, Liad Yosef and Ido Salomon of MCP Apps have detailed their vision for MCP Apps, a new standard for building interactive UI components that can be seamlessly integrated into chat interfaces. This innovation promises to redefine how users engage with applications, moving beyond static text to dynamic, interactive experiences within AI conversations.

MCP Apps: Building the Future of AI Interactions - AI Engineer
MCP Apps: Building the Future of AI Interactions — from AI Engineer

The MCP Apps Vision: Bridging UI and AI

Yosef and Salomon explained that while current AI interactions often rely on text-based responses, this approach is suboptimal for complex tasks. MCP Apps aim to solve this by enabling applications to send their own UI elements directly to AI agents. This means that instead of receiving a block of text describing a product or a data visualization, users can interact with rich, dynamic components rendered directly within their chat interface.

Related startups

Key Concepts: UI over MCP and Interoperability

The core of MCP Apps lies in its ability to transmit UI over the MCP protocol, standardizing UI-to-host communication. This allows for a bi-directional flow where an AI agent can request and render specific UI components from an application, and conversely, the UI can send actions and events back to the model. This is achieved through a system of resources and callbacks, ensuring that the host application can securely render and interact with these components.

MCP Apps are designed to be agnostic to the UI generation method, supporting a spectrum from predefined UIs (like those used by Airbnb or Spotify) to more dynamic, declarative UIs (like JSON renderers) and even fully generative UIs as demonstrated by recent advancements in the field. This flexibility ensures that existing applications can leverage the MCP standard without a complete overhaul.

The "New Web" and Future of Application Distribution

The speakers emphasized that MCP Apps are not just a technical protocol but a new way of distributing applications. They envision a future where applications are no longer confined to traditional websites or separate apps but can be seamlessly integrated into AI interfaces. This allows for a more personalized and efficient user experience, where AI agents can leverage specific UI elements from various services to fulfill user requests.

The adoption of MCP Apps is growing rapidly, with major companies like Microsoft, Google, and GitHub already integrating or supporting the standard. The rapid adoption highlights the potential for MCP Apps to become a global standard for UI within chat applications, enabling a more unified and interactive digital experience.

Getting Started and Community Involvement

For developers looking to build their own MCP Apps, the MCP Apps team provides official SDKs and a clear specification. They encourage community involvement through their GitHub repository and Discord server, where developers can contribute to the ongoing development, share tips, and collaborate on new features. The goal is to foster a robust ecosystem where applications can be easily built and deployed across various AI platforms, truly embodying the principle of "write once, run everywhere."

Ultimately, MCP Apps represents a significant step towards a more integrated and intelligent future for how we interact with technology, making AI assistants more capable and user experiences more seamless and powerful.

© 2026 StartupHub.ai. All rights reserved. Do not enter, scrape, copy, reproduce, or republish this article in whole or in part. Use as input to AI training, fine-tuning, retrieval-augmented generation, or any machine-learning system is prohibited without written license. Substantially-similar derivative works will be pursued to the fullest extent of applicable copyright, database, and computer-misuse laws. See our terms.