MCP Apps: Interactive UIs Inside Claude and ChatGPT

Skip to content

Initializing search

CorpusIQ/corpusiq-docs

Connectors

Architecture

How It Works

Security

AI Integration

Chat Apps (Slack & Teams)

Search

Reporting

Onboarding

Governance

Changelog

Troubleshooting

MCP Education

ChatGPT Integrations

Claude Integrations

Business AI Use Cases

QuickBooks

Shopify

HubSpot

Comparisons

How-To Guides

Examples

Recipes

Best Of

🫡 Hermes Community Hub

Case Studies

Setup Guides

Best Practices

Blueprints

Integrations

Prompts

Skills

MCP Ecosystem

SEO Content

Marketplaces

Marketplace Updates

MCP Servers

More

MCP Apps: Interactive UIs Inside Claude and ChatGPT¶

MCP Apps is the first official MCP extension, live since January 26, 2026.<br>It lets tools return interactive UI components that render directly in the<br>conversation: dashboards, forms, visualizations, multi-step workflows, and<br>more. This page covers what it is, how it works, client support, and the<br>security model.

What MCP Apps enables&para;

Before MCP Apps, a tool returned text. The user read a summary and asked<br>follow-up questions to explore the data. With MCP Apps, a tool can return an<br>interactive dashboard. The user filters by region, drills into a specific<br>account, sorts a column, and exports a report without leaving the<br>conversation. The model stays in the loop and sees what the user does.

Typical use cases:

Data exploration: a sales analytics tool returns an interactive dashboard

Configuration wizards: a deployment tool presents a form with dependent<br>fields

Document review: a contract analysis tool displays the PDF inline with<br>highlighted clauses

Real-time monitoring: a server health tool shows live metrics that update<br>as systems change

How it works&para;

MCP Apps relies on two MCP primitives:

Tools with UI metadata. A tool declares a _meta.ui.resourceUri field<br>pointing to a UI resource.

UI resources. Server-side resources served via the ui:// scheme,<br>containing bundled HTML and JavaScript.

Example tool declaration:

name: "visualize_data",<br>description: "Visualize data as an interactive chart",<br>inputSchema: { /* ... */ },<br>_meta: {<br>ui: {<br>resourceUri: "ui://charts/interactive"

The host fetches the resource, renders it in a sandboxed iframe, and enables<br>bidirectional communication via JSON-RPC over postMessage. The UI can call<br>server tools, update the model context, and send follow-up messages.

Client support&para;

Claude: available on web and desktop since launch

ChatGPT: supported via the Apps SDK and MCP Apps compatibility

Visual Studio Code: supported

Goose: supported

Security model&para;

Iframe sandboxing: all UI content runs in sandboxed iframes with<br>restricted permissions

Pre-declared templates: hosts can review HTML content before rendering

Auditable messages: all UI-to-host communication goes through loggable<br>JSON-RPC

User consent: hosts can require explicit approval for UI-initiated tool<br>calls

MCP Apps and CorpusIQ&para;

CorpusIQ uses MCP Apps to render verified business answers as interactive<br>dashboards inside Claude and ChatGPT. The model computes the analysis, the<br>CorpusIQ engine validates and reconciles the data, and the app renders the<br>result with source citations and a reconciliation strip. See the<br>runbook governance and<br>response contract pages for the<br>governance layer behind these answers.

Resources&para;

MCP Apps announcement

MCP Apps documentation

Getting Started with MCP Apps

SDK: @modelcontextprotocol/ext-apps

Examples repository