← back
MCP UI: Extending the frontier — Liad Yosef and Ido Salomon, MCP Apps
Takeaway
MCP Apps standardizes returning interactive, branded UI from MCP tools, decomposing the web into agent-composable atoms that keep the host model in control.
Summary
- MCPUI lets MCP tools return interactive UI (HTML resources) instead of text, now standardized via the MCP Apps spec with Anthropic and OpenAI
- Adopted by ChatGPT, Claude, VS Code, Cursor, GitHub Copilot, Postman, Goose; Shopify and Hugging Face shipped MCPUI components a year ago
- Architecture: server returns a UI resource, host renders in a sandbox, bidirectional message passing keeps the LLM in the loop on user interactions
- Enables generative-UI overlays (Claude generating explanations on top of MCPUI), preserving brand identity instead of reducing tool output to a wall of text
mcpuigenerative-ui
Original description
Chat interfaces are no longer limited to walls of text. In this talk, Liad Yosef and Ido Salomon explain how MCP Apps turn tools into interactive UI inside hosts like ChatGPT, Claude, VS Code, Cursor, and Copilot, letting companies send branded, functional app experiences instead of plain text responses. The session covers the core architecture behind MCP Apps, how UI is passed over MCP, how interactions stay in context through the host, and why this changes how applications get distributed in an agent-first world. If you're building on MCP, this is a practical look at the emerging standard for UI inside chat. Timestamps: 0:00 Introduction to MCP Apps 1:02 Why we need MCP Apps: Moving beyond text-based chat 2:06 Evolution of the MCP UI standard and partnerships 3:25 Industry and community adoption 5:14 Core concepts: Passing UI over MCP 6:49 Practical demonstration (PostHog and Claude) 8:54 Technical architecture: How it works 10:23 A new era of web interaction and user experience 12:32 Interaction mindset and message spectrums 14:56 Future outlook: Reusable views and model-UI interaction 16:18 Spectrum of UI generation: Predefined vs. Generative