What Is an MCP App?

An MCP App is an extension that renders a visual, interactive UI directly inside your AI conversation. Instead of getting a wall of text or a code block you need to copy-paste somewhere else, you see the result inline — a chart, a form, a map, a diagram — right where you asked for it.

MCP stands for Model Context Protocol, the open standard that connects AI assistants to external tools and data sources. MCP Apps build on this standard by adding a UI layer: the AI calls the tool, and the tool renders something you can see and interact with.

MCP tools vs. MCP Apps

A regular MCP tool returns text. Ask it to query a database and you get rows of data as plain text. Useful, but limited.

An MCP App returns a visual UI. Ask it to chart that same data and you get an interactive chart — with tooltips, zoom, and export — rendered right in the conversation. The difference is like reading a spreadsheet vs. seeing a dashboard.

How MCP Apps work

Under the hood, an MCP App is an MCP server that registers both tools and a UI resource. When the AI calls a tool, the server returns structured data along with a reference to its UI. The host application (Claude, ChatGPT, VS Code, etc.) loads the UI and passes the data to it. The UI renders client-side — no extra servers, no iframes pointing to third-party domains.

This means your data stays local. The UI runs in your client, and the rendering happens in your browser. Nothing is sent to an external service beyond what you already shared with your AI assistant.

Where MCP Apps work

MCP Apps work in any client that supports the MCP Apps extension. As of early 2026, that includes:

• Claude and Claude Desktop
• ChatGPT
• VS Code and Cursor
• Goose, Postman, MCPJam, and others

You install an MCP App once, and it works across all these clients. No per-client plugins or extensions to manage.

Examples of MCP Apps

MCP Apps can render anything that makes sense as a visual response:

• Charts and dashboards — bar, line, pie, scatter, radar charts from data in your conversation
• Maps — plot locations, routes, and geospatial data
• Diagrams — flowcharts, architecture diagrams, entity relationships
• Forms and controls — interactive inputs that feed back into the conversation

ChartPane is an MCP App for data visualization. Ask your AI assistant to chart something and you get interactive, production-quality charts rendered inline — with tooltips, legends, and one-click PNG export.

Try it yourself

The fastest way to experience an MCP App is to install ChartPane. It takes about a minute: add a single URL to your AI client's MCP settings, and you're done. No API keys, no sign-up, no cost.