OpenAI Apps MCP

Deploy stateless MCP servers on Cloudflare Workers for ChatGPT extensions.

961|99|Updated Oct 20, 2025
One-click install
npx skills add https://github.com/jezweb/claude-skills --skill openai-apps-mcp
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: OpenAI Apps MCP
Source: https://github.com/jezweb/claude-skills/tree/main/skills/openai-apps-mcp
Command: npx skills add https://github.com/jezweb/claude-skills --skill openai-apps-mcp

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @modelcontextprotocol/sdk, hono, zod, @cloudflare/vite-plugin, vite, wrangler, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill eliminates the complexity of integrating MCP servers with ChatGPT, preventing common errors like CORS blocking, widget 404s, and MIME type issues that typically derail development.

Core Features & Use Cases

  • MCP Server Setup: Create stateless MCP servers using Hono and Cloudflare Workers.
  • Widget Integration: Serve interactive HTML/JS widgets that render directly in ChatGPT conversations.
  • Use Case: Imagine you want to create a restaurant finder app for ChatGPT. Use this Skill to build an MCP server that shows restaurant locations on an interactive map widget.

Quick Start

Use the OpenAI Apps MCP skill to create a ChatGPT app that displays a greeting widget when users ask to say hello.

Frequently Asked Questions about OpenAI Apps MCP

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a ChatGPT app that uses custom tools and widgets?

Build ChatGPT apps by deploying stateless MCP servers on Cloudflare Workers using the @modelcontextprotocol/sdk and Hono. Your MCP server extends ChatGPT with custom tools and interactive HTML/JS widgets that render directly in conversations.

What is Model Context Protocol and why use it with ChatGPT?

Model Context Protocol (MCP) is a standardized way to connect ChatGPT with external tools and data sources. It lets you define custom tools that ChatGPT can invoke and serve interactive widgets alongside responses without rebuilding ChatGPT itself.

Can I serve interactive widgets in ChatGPT conversations?

Yes. Deploy your widgets as assets on Cloudflare Workers and reference them in MCP tool definitions using correct MIME types and CORS configuration. Widgets render natively in ChatGPT when your MCP server provides the proper outputTemplate and widget URIs.

Why am I getting widget 404 errors or CORS issues with my ChatGPT integration?

Widget 404s and CORS failures typically stem from misconfigured asset delivery on Cloudflare Workers or incorrect widget URIs in your MCP server metadata. Verify CORS headers, MIME types, and that Cloudflare Workers is serving your widget files at the declared URIs.

Do I need Cloudflare Workers to run an MCP server for ChatGPT?

This Skill uses Cloudflare Workers as the deployment platform for stateless MCP servers because they provide serverless execution and integrated asset serving. You configure your MCP server with Hono, Vite, and Wrangler for seamless deployment.

What's the best way to test ChatGPT tool invocations and widget rendering locally?

Deploy your MCP server to Cloudflare Workers, configure it in ChatGPT's tools settings, then test tool invocations and widget rendering through ChatGPT's interface. The Skill covers troubleshooting SSE heartbeat, MIME types, and asset-based widget delivery for end-to-end validation.