Cover Image for [officehours 44] MCP Apps: Enhancing Agents with Interactive Apps
Cover Image for [officehours 44] MCP Apps: Enhancing Agents with Interactive Apps
Avatar for intelcraft
Presented by
intelcraft
AI Business Design, Engineering, Architecture with NoCode/AI
17 Going
Hayden
invites you to join

[officehours 44] MCP Apps: Enhancing Agents with Interactive Apps

Google Meet
Registration
Welcome! To join the event, please register below.
About Event

In Office Hours 43, we covered cybersecurity and looked at how agents can reason across security workflows, tools, and data.

For Office Hours 44, we’re going to look at MCP Apps, the new official extension to the Model Context Protocol that lets MCP tools return interactive user interfaces directly inside an agent experience — things like dashboards, charts, forms, file uploaders, design canvases, and multi-step workflows. In other words: MCP is moving beyond “call a tool, get JSON back” toward “call a tool, open an app.”

This matters because agents are increasingly becoming the place where work happens. Instead of jumping between Slack, Figma, Canva, Asana, dashboards, and internal tools, MCP Apps make it possible for those interfaces to appear inside the conversation itself. Anthropic has already shown this pattern with apps like Slack, Figma, Canva, and Asana opening interactively inside Claude.

We’ll focus on the developer side: how these apps are built, what libraries are emerging, and how different ecosystems are approaching the problem.

We’ll cover examples including:

  1. Prefab + FastMCP for Python
    Prefab is a generative UI framework for MCP Apps, designed as a flagship use case for FastMCP. It lets Python developers compose interactive UI components and deliver them through MCP Apps without having to hand-build a full frontend stack.

  2. mcp-ui for Node / TypeScript
    mcp-ui is an SDK implementing the MCP Apps standard for UI over MCP, with TypeScript server support for creating UI resources. This is one of the more direct Node/TypeScript analogs to the “UI over MCP” pattern.

  3. OpenAI Apps SDK for MCP-backed apps
    OpenAI’s Apps SDK uses MCP servers, tools, resources, and inline HTML/widgets to render interactive app experiences in ChatGPT. It’s a useful reference point for how MCP-style tools become rich conversational apps.

  4. C# / .NET MCP SDK and emerging Apps support
    The official C# SDK enables .NET applications to implement MCP clients and servers, and there is active work/discussion around adding an Apps convenience layer similar to the TypeScript extension approach. We’ll look at what’s available now, what’s still emerging, and what .NET teams can already build.

The big question for this session: what changes when an MCP server doesn’t just expose a tool, but exposes an app?

We’ll talk through what this means for internal tools, enterprise workflows, agent UX, and the next generation of software interfaces.

And as always, if you’re building in this space and want 10–15 minutes to show off your workflow, library, or MCP App experiment, let me know and I’ll add you to the program.

Avatar for intelcraft
Presented by
intelcraft
AI Business Design, Engineering, Architecture with NoCode/AI
17 Going