MCP Apps

MCP Apps + CopilotKit

Build MCP-powered interactions directly into your agentic application.

MCP Apps were initially designed to run inside chat-based AI hosts such as ChatGPT and Claude.

CopilotKit and AG-UI now make MCP Apps usable inside real, user-facing applications.

What are MCP Apps?

MCP Apps are declared by MCP servers and represent distinct user interactions — not just function calls.

Each MCP App corresponds to a purpose-built interaction surface that a host can render and interact with. A single MCP server may expose multiple MCP Apps, each optimized for a different user task.

•MCP Apps represent interactive capabilities

•They are open-ended by design (UI is not prescribed by the spec)

•They allow tools to bring non-text interaction into agent workflows

“MCP Apps are a game-changer. They join protocols like MCP, A2A, AG-UI and Generative UI specs to unlock unprecedented composability and UX. This is what we built CopilotKit for: connecting any app to any agentic backend. With MCP Apps, that vision just got supercharged.”

Atai Barkai

CEO and Co-Founder, CopilotKit

CopilotKit: Your App + MCP Apps

CopilotKit and AG-UI bring MCP apps into your application — where you control UX, permissions, workflows, and agent behavior — instead of confining them to first-party AI chat experiences.

Using MCP Apps through CopilotKit lets your app move beyond text-only copilots and become an interactive, agent-driven product where tools can ask users for input, confirmation, and decisions at the exact moment they matter.

Powered by MCP Apps and CopilotKit, your application delivers a unified experience where text, tools, UI, and user input work together seamlessly.

Tool-Native UI

Let agents open pickers, forms, and approvals instead of asking users to respond in text

Structured Input

Collect structured input that agents can reliably act on

In-Product User Retention

Keep users inside your product, rather than bouncing them to external chat tools

Workflow Integrated Actions

Blend agent actions directly into existing workflows and screens

The Agentic App Stack using MCP Apps

CopilotKitHandles discovery, renders UI, manages user interaction, and integrates results back into the application

AG-UIManages all communication between agentic backend and user-facing application. Enables interaction flow end-to-end.

AI AgentsDecide when a user needs to be involved and which MCP App to invoke.

MCP ServersDeclare and expose MCP Apps as capabilities for agents to invoke.

MCP AppsDefine interactive experiences that are delivered to users within agentic apps

Are you ready?

Ready to Get Started?

Learn how to integrate MCP Apps into your application with CopilotKit and AG-UI.