Hello, Plugins! — Python OpenAI Plugin

This is a minimal OpenAI Plugin example that returns exactly Hello, Plugins! when the hello_plugins tool is called.

An OpenAI Plugin can be composed of a Skill, an MCP server, or both. This project is the smallest form that requires executable code, so it uses only one Python MCP server. Since it does not call the OpenAI API directly, OPENAI_API_KEY is not required either.

Project Structure

hello-plugins-python/
├── server.py          # MCP 서버와 hello_plugins 도구
├── test_server.py     # 반환 문자열 테스트
├── requirements.txt  # 고정된 Python 의존성
├── pyproject.toml
├── Dockerfile
└── README.md

Related MCP server: hello-mcp

Running on Windows 11 / PowerShell

Python 3.10 or later is required.

cd hello-plugins-python
py -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install -r requirements.txt
python server.py

After running, the MCP Streamable HTTP address is as follows:

http://localhost:8000/mcp

If the PowerShell execution policy blocks activation of the virtual environment, first run the following command in the current terminal only.

Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass

Running on macOS / Linux

cd hello-plugins-python
python3 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt
python server.py

Testing

python -m pytest -q
python verify_mcp.py

Expected result:

1 passed
MCP verification passed: Hello, Plugins!

Verifying with MCP Inspector

If Node.js is installed, run the following:

npx @modelcontextprotocol/inspector

In the Inspector, select Streamable HTTP for Transport and enter http://localhost:8000/mcp as the URL. After connecting, calling the hello_plugins tool returns the following string:

Hello, Plugins!

Connecting a Personal Plugin to ChatGPT Work

ChatGPT cannot directly access localhost on a local PC, so to test it you need to deploy the /mcp endpoint to a public HTTPS URL or use the OpenAI Secure MCP Tunnel.

1. Deploy the server to a public HTTPS environment.

2. In ChatGPT, turn on Settings → Security and login → Developer mode.

3. On the Plugins screen, press + and enter https://YOUR-DOMAIN/mcp.

4. Install the generated Personal Plugins.

5. In a Work chat, select @Hello Plugins and ask, "Print the plugin greeting."

Running in Docker

docker build -t hello-plugins .
docker run --rm -p 8000:8000 hello-plugins

Core Code

@mcp.tool(
    name="hello_plugins",
    description="Return the exact greeting 'Hello, Plugins!'. This tool is read-only.",
)
def hello_plugins() -> str:
    return "Hello, Plugins!"

References

• Official OpenAI Plugin documentation: https://developers.openai.com/plugins

• Plugin Quickstart: https://developers.openai.com/plugins/quickstart

• MCP server development: https://developers.openai.com/plugins/build/mcp-server