If you've been using AI assistants for a while, you've probably hit the ceiling of what plain text responses can do. Ask an AI to analyse your sales data and you get a paragraph. Ask it to build a dashboard and you get Python code you have to run yourself.
MCP Apps change that equation.
What Are MCP Apps?
MCP Apps are a special type of Model Context Protocol extension that returns interactive HTML interfaces instead of plain text. When an AI agent calls an MCP App, the response renders as a live UI component — a chart, a form, a configuration panel, a data table — directly inside the AI client's conversation.
The AI client handles the rendering. You get a rich, interactive experience without leaving the chat and without the AI having to output code you then have to execute.
This isn't a gimmick. For workflows that naturally produce structured data — analytics, configuration, data entry, progress tracking — interactive UI beats prose every time.
How MCP Apps Work
MCP Apps are built as MCP Servers that implement the MCP Apps protocol extension. The server exposes tools that return UI definitions rather than text. The MCP host (Claude Desktop, VS Code Copilot, ChatGPT, Goose) renders the returned HTML inline.
For developers, building an MCP App is straightforward:
// MCP App server example (simplified)
export const mcpApp = {
name: "sales-dashboard",
tools: {
showDashboard: {
type: "app",
content: {
component: "Dashboard",
props: { title: "Q1 Sales", data: "{{sales_data}}" }
}
}
}
}
For users, the experience is seamless — you ask for the dashboard, you get a dashboard.
Where MCP Apps Work
Currently supported by:
- Claude Desktop (Anthropic)
- VS Code Copilot (Microsoft)
- ChatGPT (OpenAI)
- Goose (Block)
- Any MCP-compatible host that implements the UI extension spec
What You Can Build With MCP Apps
The space is moving fast. Developers are building:
- Data dashboards — query a database and get a live chart
- Configuration UIs — generate and visualise infrastructure configs
- Form wizards — multi-step onboarding and data collection flows
- Real-time monitors — live logs, streaming status, progress bars
- Media viewers — 3D model previews, video players, waveform visualisers
Browse the MCP Apps Directory
We track every MCP App in one place — with descriptions, compatibility info, install commands, and links to source code.
New MCP Apps are discovered and added daily by our crawling system. If you're building one, submit it here.