Joke Button MCP Server
A minimal MCP server that exposes a single-button UI. Clicking the button prompts the assistant to deliver a short, family-friendly joke.
README
MCP Joke Button App
This project is a tiny MCP server that exposes one MCP App UI: a single button. When the button is clicked, the app sends a message to the host asking the assistant to answer with one short, family-friendly joke.
It intentionally uses only Node.js built-in modules. There are no runtime package dependencies.
What It Demonstrates
- MCP core protocol version
2026-07-28 - Stateless Streamable HTTP on one endpoint:
/mcp server/discover,tools/list,tools/call,resources/list, andresources/read- MCP Apps UI resource metadata with
_meta.ui.resourceUri - A sandbox-friendly
text/html;profile=mcp-appwidget ui/initializeandui/messageoverpostMessage
Requirements
Install Node.js 20 or newer from nodejs.org. The LTS download is fine.
You do not need to run npm install; this project has no third-party libraries.
Start the Server
Open a terminal in this folder and run:
npm start
You should see:
MCP Joke Button server listening on http://127.0.0.1:8787/mcp
Preview the widget shell at http://127.0.0.1:8787/preview
Leave that terminal open while you test.
MCP requests are logged by method while the server is running. For an interactive app render, the log should include a successful UI resource read similar to:
[mcp] request method=resources/read name=ui://joke-button/app.html
[mcp] response method=resources/read status=200
Set MCP_LOG_REQUESTS=false if you want to disable these development logs.
Heroku Deployment
The joke-button Heroku app is deployed at:
https://joke-button-31cddc4bc964.herokuapp.com
Its MCP endpoint is:
https://joke-button-31cddc4bc964.herokuapp.com/mcp
Deploy the current branch with:
git push heroku HEAD:main
Quick Local Check
Open this address in a browser:
http://127.0.0.1:8787/preview
This only previews the button's appearance. The button can ask the AI for a joke only when it is rendered inside an MCP Apps-compatible host such as ChatGPT.
Use with ChatGPT Desktop or ChatGPT Web
ChatGPT needs an HTTPS URL for a remote MCP server during development. Keep npm start running, then expose port 8787 with a tunnel tool such as ngrok or Cloudflare Tunnel.
Example with ngrok:
ngrok http 8787
Copy the HTTPS forwarding URL and add /mcp to the end, for example:
https://example.ngrok.app/mcp
Then connect it in ChatGPT:
- Open ChatGPT.
- Open Settings.
- Open Security and login.
- Turn on Developer mode if it is not already enabled.
- Open Plugins. If you are in the desktop app and do not see it in the sidebar, open
https://chatgpt.com/pluginsin a browser while signed in to the same account. - Select the plus button to add a developer MCP server connection.
- Paste the HTTPS URL ending in
/mcp. - Name it
Joke Button. - Start a new chat, select the plugin from the More menu, and ask:
Show the joke button. - Click
Tell a joke.
You do not need a setting named MCP servers for this ChatGPT UI test. That setting is for Codex's local MCP host configuration, not for adding a ChatGPT developer plugin connection.
If Developer mode or the plus button is missing, check that the desktop app is up to date and that your account or workspace allows plugin development.
If the app is working, ChatGPT should receive the button's message and respond with a short joke.
Use with Codex in the ChatGPT Desktop App
Codex can connect to Streamable HTTP MCP servers from the shared MCP settings.
The MCP servers setting appears only on Codex-capable local clients and may not appear in a regular ChatGPT chat settings view. If it is available:
- Keep
npm startrunning. - Open the ChatGPT desktop app.
- Open Settings, then MCP servers.
- Select Add server.
- Choose Streamable HTTP.
- Name the server
joke_button. - Use this URL:
http://127.0.0.1:8787/mcp
- Save, then restart when prompted.
- In a Codex task, type
/mcpto confirm the server is connected. - Ask Codex:
Use the Joke Button MCP server.
Codex MCP tool support may show the text fallback rather than rendering the iframe UI, depending on the surface. For the actual one-button UI, use ChatGPT with MCP Apps UI support.
If the setting is not visible, copy the sample configuration from codex-config.example.toml into a trusted project's .codex/config.toml or into your user ~/.codex/config.toml, then restart the ChatGPT desktop app or Codex client.
Useful URLs
- Health check:
http://127.0.0.1:8787/health - MCP endpoint:
http://127.0.0.1:8787/mcp - Widget preview:
http://127.0.0.1:8787/preview
Security Notes
The server validates the MCP protocol version and mirrored MCP headers required by the 2026-07-28 Streamable HTTP transport.
For local development, requests without an Origin header are accepted, loopback browser origins are accepted, and common OpenAI host origins are accepted. For another host origin, start the server with:
$env:MCP_ALLOWED_ORIGINS="https://your-host.example"
npm start
Use authentication before exposing a real MCP server that can access private data or perform actions. This sample does neither.
Project Files
server.js: dependency-free MCP serverpublic/joke-button.html: MCP App UI widgetcodex-config.example.toml: sample Codex MCP configuration
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.