Validate ChatGPT App

You are helping the user validate their ChatGPT App before testing and deployment.

Validation Checks

1. Required Files Check (RUN FIRST)

Check that ALL mandatory files exist:

ls package.json tsconfig.server.json setup.sh START.sh .env.example server/index.ts

Expected file structure:

{app-name}/
├── package.json
├── tsconfig.server.json
├── setup.sh
├── START.sh
├── .env.example
├── .gitignore
└── server/
    └── index.ts

If ANY of the above are missing, report as CRITICAL ERROR.

2. Server Implementation Check

Verify server/index.ts uses correct patterns:

MUST have:

import { Server } from "@modelcontextprotocol/sdk/server/index.js";
import { StreamableHTTPServerTransport } from "@modelcontextprotocol/sdk/server/streamableHttp.js";

MUST NOT have:

import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";  // WRONG

Session management MUST exist:

const transports = new Map<string, StreamableHTTPServerTransport>();

3. Widget Configuration Check

Verify widgets array structure:

const widgets: WidgetConfig[] = [
  {
    id: "widget-id",                           // kebab-case
    name: "Widget Name",
    description: "What it displays",
    templateUri: "ui://widget/widget-id.html", // MUST match this format
    invoking: "Loading...",
    invoked: "Ready",
    mockData: { /* sample data */ },
  },
];

4. Tool Response Check

Widget tools MUST return:

return {
  content: [{ type: "text", text: JSON.stringify(result) }],
  structuredContent: result,  // CRITICAL - becomes window.openai.toolOutput
  _meta: {
    "openai/outputTemplate": widget.templateUri,
    "openai/toolInvocation/invoked": widget.invoked,
  },
};

5. Resource Handler Check

Verify ReadResource returns correct format:

return {
  contents: [{
    uri,
    mimeType: "text/html+skybridge",  // MUST be this exact value
    text: generateWidgetHtml(widget.id),
  }],
  _meta: {
    "openai/serialization": "markdown-encoded-html",
    "openai/csp": { ... },
  },
};

6. Widget HTML Check

Verify generateWidgetHtml includes:

• window.PREVIEW_DATA

  support for local preview

• window.openai.toolOutput

  data access

• openai:set_globals

  event listener
• Polling fallback with setInterval

• rendered

  flag to prevent re-renders

7. Endpoint Check

Verify these endpoints exist:

• GET /health

  - Health check

• GET /preview

  - Widget preview index

• GET /preview/:widgetId

  - Individual widget preview

• ALL /mcp

  - MCP protocol handler

• DELETE /mcp

  - Session cleanup

8. Package.json Scripts Check

{
  "scripts": {
    "build": "npm run build:server",
    "build:server": "tsc -p tsconfig.server.json",
    "start": "HTTP_MODE=true node dist/server/index.js",
    "dev": "HTTP_MODE=true NODE_ENV=development tsx watch --clear-screen=false server/index.ts"
  }
}

MUST NOT have:

dev:web

build:web

concurrently

9. Annotation Validation

• Query tools have

  readOnlyHint: true

• Delete tools have

  destructiveHint: true

• External API tools have

  openWorldHint: true

10. Database Validation (if enabled)

• All migrations are valid SQL
• Tables have

  user_subject

  column
• Indexes exist for user queries

Workflow

1. Run file existence checks
2. Read and analyze server/index.ts
3. Verify patterns match requirements
4. Collect errors and warnings
5. Display results

Results Format

## Validation Results

### File Structure ✓
All required files present.

### Server Implementation ✓
Using correct Server class with session management.

### Widget Configuration ✓
2 widgets properly configured.

### Tool Responses ✓
All widget tools return structuredContent.

---
**Overall: PASS**

Common Errors

Server

@modelcontextprotocol/sdk/server/index.js

Map<string, StreamableHTTPServerTransport>

ui://widget/{id}.html

text/html+skybridge