json-render/apps/web/app/(main)/docs/api/mcp/page.mdx

import { pageMetadata } from "@/lib/page-metadata"
export const metadata = pageMetadata("docs/api/mcp")

# @json-render/mcp

MCP Apps integration for json-render. Serve json-render UIs as interactive [MCP Apps](https://modelcontextprotocol.io/docs/extensions/apps) inside Claude, ChatGPT, Cursor, VS Code, and other MCP-capable clients.

## Install

```bash
npm install @json-render/mcp @json-render/core @modelcontextprotocol/sdk
```

For the iframe-side React UI, also install:

```bash
npm install @json-render/react react react-dom
```

See the [MCP example](https://github.com/vercel-labs/json-render/tree/main/examples/mcp) for a full working example.

## Overview

MCP Apps let MCP servers return interactive HTML UIs that render directly inside chat conversations. `@json-render/mcp` bridges json-render catalogs with the MCP Apps protocol:

1. Your **catalog** defines which components and actions the AI can use
2. The **MCP server** exposes the catalog as a tool with the spec schema
3. The **bundled HTML** renders json-render specs inside the host's sandboxed iframe
4. The AI generates a spec, the host renders it, and users interact with the live UI

## Server API

### createMcpApp

Create a fully-configured MCP server. This is the main entry point.

```typescript
import { createMcpApp } from "@json-render/mcp";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
import fs from "node:fs";

const server = createMcpApp({
  name: "My Dashboard",
  version: "1.0.0",
  catalog: myCatalog,
  html: fs.readFileSync("dist/index.html", "utf-8"),
});

await server.connect(new StdioServerTransport());
```

#### CreateMcpAppOptions

<table>
  <thead>
    <tr>
      <th>Option</th>
      <th>Type</th>
      <th>Description</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td><code>name</code></td>
      <td><code>string</code></td>
      <td>Server name shown in client UIs</td>
    </tr>
    <tr>
      <td><code>version</code></td>
      <td><code>string</code></td>
      <td>Server version</td>
    </tr>
    <tr>
      <td><code>catalog</code></td>
      <td><code>Catalog</code></td>
      <td>json-render catalog defining available components</td>
    </tr>
    <tr>
      <td><code>html</code></td>
      <td><code>string</code></td>
      <td>Self-contained HTML for the iframe UI</td>
    </tr>
    <tr>
      <td><code>tool</code></td>
      <td><code>McpToolOptions</code></td>
      <td>Optional tool name/title/description overrides</td>
    </tr>
  </tbody>
</table>

### registerJsonRenderTool

Register a json-render tool on an existing `McpServer`. Use this when you need to add json-render to a server that has other tools.

```typescript
import { registerJsonRenderTool } from "@json-render/mcp";

registerJsonRenderTool(server, {
  catalog,
  name: "render-ui",
  title: "Render UI",
  description: "Render an interactive UI",
  resourceUri: "ui://render-ui/view.html",
});
```

### registerJsonRenderResource

Register the UI resource that serves the bundled HTML.

```typescript
import { registerJsonRenderResource } from "@json-render/mcp";

registerJsonRenderResource(server, {
  resourceUri: "ui://render-ui/view.html",
  html: bundledHtml,
});
```

## Client API (`@json-render/mcp/app`)

These exports run inside the sandboxed iframe rendered by the MCP host.

### useJsonRenderApp

React hook that connects to the MCP host, listens for tool results, and maintains the current json-render spec.

```tsx
import { useJsonRenderApp } from "@json-render/mcp/app";
import { JSONUIProvider, Renderer } from "@json-render/react";

function McpAppView({ registry }) {
  const { spec, loading, connected, error } = useJsonRenderApp({
    name: "my-app",
    version: "1.0.0",
  });

  if (error) return <div>Error: {error.message}</div>;
  if (!spec) return <div>Waiting...</div>;

  return (
    <JSONUIProvider registry={registry} initialState={spec.state ?? {}}>
      <Renderer spec={spec} registry={registry} loading={loading} />
    </JSONUIProvider>
  );
}
```

#### UseJsonRenderAppReturn

<table>
  <thead>
    <tr>
      <th>Field</th>
      <th>Type</th>
      <th>Description</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td><code>spec</code></td>
      <td><code>{'Spec | null'}</code></td>
      <td>Current json-render spec</td>
    </tr>
    <tr>
      <td><code>loading</code></td>
      <td><code>boolean</code></td>
      <td>Whether the spec is still being received</td>
    </tr>
    <tr>
      <td><code>connected</code></td>
      <td><code>boolean</code></td>
      <td>Whether connected to the host</td>
    </tr>
    <tr>
      <td><code>connecting</code></td>
      <td><code>boolean</code></td>
      <td>Whether currently connecting</td>
    </tr>
    <tr>
      <td><code>error</code></td>
      <td><code>{'Error | null'}</code></td>
      <td>Connection error, if any</td>
    </tr>
    <tr>
      <td><code>app</code></td>
      <td><code>{'App | null'}</code></td>
      <td>The underlying MCP App instance</td>
    </tr>
    <tr>
      <td><code>callServerTool</code></td>
      <td><code>{'(name, args?) => Promise<void>'}</code></td>
      <td>Call an MCP server tool and update spec from result</td>
    </tr>
  </tbody>
</table>

### buildAppHtml

Generate a self-contained HTML page from bundled JavaScript and CSS.

```typescript
import { buildAppHtml } from "@json-render/mcp/app";
import fs from "node:fs";

const html = buildAppHtml({
  title: "Dashboard",
  js: fs.readFileSync("dist/app.js", "utf-8"),
  css: fs.readFileSync("dist/app.css", "utf-8"),
});
```

## Client Configuration

### Cursor

Add to `.cursor/mcp.json`:

```json
{
  "mcpServers": {
    "json-render": {
      "command": "npx",
      "args": ["tsx", "path/to/server.ts", "--stdio"]
    }
  }
}
```

### Claude Desktop

Add to `~/Library/Application Support/Claude/claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "json-render": {
      "command": "npx",
      "args": ["tsx", "/absolute/path/to/server.ts", "--stdio"]
    }
  }
}
```

## Supported Clients

MCP Apps are supported by Claude (web and desktop), ChatGPT, VS Code (GitHub Copilot), Cursor, Goose, and Postman.