Repository files navigation

This repository showcases example UI components to be used with the Apps SDK, as well as example MCP servers that expose a collection of components as tools.It is meant to be used as a starting point and source of inspiration to build your own apps for ChatGPT.

The Model Context Protocol (MCP) is an open specification for connecting large language model clients to external tools, data, and user interfaces. An MCP server exposes tools that a model can call during a conversation and returns results according to the tool contracts. Those results can include extra metadata—such as inline HTML—that the Apps SDK uses to render rich UI components (widgets) alongside assistant messages.

Within the Apps SDK, MCP keeps the server, model, and UI in sync. By standardizing the wire format, authentication, and metadata, it lets ChatGPT reason about your connector the same way it reasons about built-in tools. A minimal MCP integration for Apps SDK implements three capabilities:

1. List tools – Your server advertises the tools it supports, including their JSON Schema input/output contracts and optional annotations (for example,readOnlyHint).
2. Call tools – When a model selects a tool, it issues acall_tool request with arguments that match the user intent. Your server executes the action and returns structured content the model can parse.
3. Return widgets – Alongside structured content, return embedded resources in the response metadata so the Apps SDK can render the interface inline in the Apps SDK client (ChatGPT).

Because the protocol is transport agnostic, you can host the server over Server-Sent Events or streaming HTTP—Apps SDK supports both.

The MCP servers in this demo highlight how each tool can light up widgets by combining structured payloads with_meta.openai/outputTemplate metadata returned from the MCP servers.

• src/ – Source for each widget example.
• assets/ – Generated HTML, JS, and CSS bundles after running the build step.
• pizzaz_server_node/ – MCP server implemented with the official TypeScript SDK.
• pizzaz_server_python/ – Python MCP server that returns the Pizzaz widgets.
• solar-system_server_python/ – Python MCP server for the 3D solar system widget.
• build-all.mts – Vite build orchestrator that produces hashed bundles for every widget entrypoint.

• Node.js 18+
• pnpm (recommended) or npm/yarn
• Python 3.10+ (for the Python MCP server)

Clone the repository and install the workspace dependencies:

pnpm install

Using npm or yarn? Install the root dependencies with your preferred client and adjust the commands below accordingly.

The components are bundled into standalone assets that the MCP servers serve as reusable UI resources.

pnpm run build

This command runsbuild-all.mts, producing versioned.html,.js, and.css files insideassets/. Each widget is wrapped with the CSS it needs so you can host the bundles directly or ship them with your own server.

To iterate locally, you can also launch the Vite dev server:

pnpm run dev

If you want to preview the generated bundles without the MCP servers, start the static file server after running a build:

pnpm run serve

The assets are exposed athttp://localhost:4444 with CORS enabled so that local tooling (including MCP inspectors) can fetch them.

The repository ships several demo MCP servers that highlight different widget bundles:

• Pizzaz (Node & Python) – pizza-inspired collection of tools and components
• Solar system (Python) – 3D solar system viewer

Every tool response includes plain text content, structured JSON, and_meta.openai/outputTemplate metadata so the Apps SDK can hydrate the matching widget.

cd pizzaz_server_nodepnpm start

python -m venv .venvsource .venv/bin/activatepip install -r pizzaz_server_python/requirements.txtuvicorn pizzaz_server_python.main:app --port 8000

python -m venv .venvsource .venv/bin/activatepip install -r solar-system_server_python/requirements.txtuvicorn solar-system_server_python.main:app --port 8000

You can reuse the same virtual environment for all Python servers—install the dependencies once and run whichever entry point you need.

To add these apps to ChatGPT, enabledeveloper mode, and add your apps in Settings > Connectors.

To add your local server without deploying it, you can use a tool likengrok to expose your local server to the internet.

For example, once your mcp servers are running, you can run:

ngrok http 8000

You will get a public URL that you can use to add your local server to ChatGPT in Settings > Connectors.

For example:https://<custom_endpoint>.ngrok-free.app/mcp

• Customize the widget data: edit the handlers inpizzaz_server_node/src,pizzaz_server_python/main.py, or the solar system server to fetch data from your systems.
• Create your own components and add them to the gallery: drop new entries intosrc/ and they will be picked up automatically by the build script.

You are welcome to open issues or submit PRs to improve this app, however, please note that we may not review all suggestions.

This project is licensed under the MIT License. SeeLICENSE for details.