How to Create a Codex Plugin
In a previous lesson, you downloaded a spreadsheet that had lots of fake data in it. This one:
How did we create this fake data? Why, with Codex, of course! Once we pointed it at a Python library called Faker. In this lesson, you'll learn how to create fake data for yourself. But, most importantly, you'll learn how to turn this fake data creation into a plugin. Then, whenever you want some data for a project, you can just use your Fake Data plugin
What Is Python Faker?
If you have a look at the following page, you'll see that it is the documentation for the Faker Python library:
https://faker.readthedocs.io/en/master/
Luckily, you don't have to read and understand all the documentation yourself - just ask Codex to absorb the knowledge.
In a new Codex project, add the following prompt:
You may want your data to be something else. In which case, just amend the final line. For example, something like this:
"I want my data be for a fake store that sells antique toys and games. I'll need fake products, fake sales data, and fake customers."
Send the prompt and Codex will get to work. If you have never installed anything to do with the Python programming language, Codex should download everything for you. It may, however, pause and ask for your permission. And it could take a while, so you'll need to be patient.
When it's done, Codex will come back with something like this:
You can then click the link to view the file. (If you left-click, Codex will try to open the file in its own inbuilt browser. You'll see a panel appear on the right. Toggle the panel by clicking the icon top right. If you right-click the link instead, you'll get a menu where you can see lots more ways to open the file.)
What Is a Codex Plugin?
If you liked how easy it was to create some fake data, and you think it might be useful in the future, you can add Faker as plugin. That way, you'll easily be able to create data anytime you like. Before we do that, though, it would be nice to know what a plugin is.
A Codex plugin is an installable package of reusable capabilities. It can contain one or more skills that provide instructions for a task, as well as connectors or MCP tools that let Codex work with external services and information. In this lesson, the plugin packages the Faker mock-data workflow so you can select and reuse it in future projects.
Plugins that need live data or actions can connect to other software through MCP servers. The MCP lesson later in this course explains how that connection works and demonstrates it with Codex and Blender.
You can see a list of all available plugins by clicking on Plugin on the left of Codex:
You'll see something like this appear in the main area:
Some of the plugins come pre-installed; for others, you have to click the install button, like the Computer Use one in the image above. The idea is that you can ask Codex to use one of your plugins to do things. For example:
"Use Spreadsheets to create a simple day planner spreadsheet."
Notice that there are tables called Public and Personal. When you create your own plugins, they will appear under the Personal tab.
Notice, too, that at the very top there is a tab for Plugins and tab for Skills. What is the difference between a plugin and a skill? A skill teaches Codex how to perform a particular task. A plugin is an installable package that can contain one or more skills, along with connections to other tools or services.
A simple way to remember it:
Skill = a recipe, instructions for completing a task.
Plugin = a toolbox, a package containing recipes and possibly extra tools.
For example, a meeting assistant skill might explain how to turn notes into actions. A meeting plugin could contain that skill plus a connection to Google Calendar or Slack.
A skill can exist by itself or be included inside a plugin. You usually create a skill while developing a workflow, then package it as a plugin when you want other people to install and use it.
(You can manage your plugins by clicking the cog icon in the top right of the main Codex window.)
Now that you what a plugin is, let's create one. It's quite easy - just ask Codex to do it!
Create a Plugin
Go back to you Chat window, the one where you asked for the fake data. Now add this prompt:
Once Codex does its thing, you should get something like this back (only with two sample prompts added):
Now go back to your Plugins page on the left of Codex. You should see your new Faker plugin there:
Notice that it has a default icon that is rather plain and grey. You can change this icon but, at time of writing, this is little fiddly to do. But let's give it a go. Here are two icons you can download:
Right-click the two icons above and save them to your own computer. Keep the file names, if you're following along.
Now you need to navigate to the plugins directory on your own computer. It will be somewhere like this file path:
C:\Users\YOUR_USER_NAME\plugins\faker-mock-data
You'll then see a folder called .codex-plugin. Inside this folder you will find a file called plugin.json. Double-click the json file to open it up and you'll some code:
{
"name": "faker-mock-data",
"version": "0.1.0",
"description": "Generate realistic, reproducible synthetic datasets with Python Faker and domain-aware business rules.",
"author": {
"name": "Personal"
},
"license": "MIT",
"keywords": [
"faker",
"mock-data",
"synthetic-data",
"test-data",
"fixtures"
],
"skills": "./skills/",
"interface": {
"displayName": "Faker Mock Data",
"shortDescription": "Generate realistic synthetic data with Faker.",
"longDescription": "Create reproducible mock datasets for prototypes, tests, demos, spreadsheets, APIs, and database seeds. Uses Faker for identity fields and explicit business rules for realistic metrics and relationships.",
"developerName": "Personal",
"category": "Productivity",
"capabilities": [
"Write"
],
"defaultPrompt": [
"Create a realistic mock customer dataset.",
"Generate synthetic sales data for Excel.",
"Make reproducible JSON fixtures for my app."
]
}
}
You need to update the interface section of the json file:
"interface": {
"displayName": "Faker Mock Data",
"shortDescription": "Generate realistic synthetic data with Faker.",
"longDescription": "Create reproducible mock datasets for prototypes, tests, demos, spreadsheets, APIs, and database seeds. Uses Faker for identity fields and explicit business rules for realistic metrics and relationships.",
"developerName": "Personal",
"category": "Productivity",
"capabilities": [
"Write"
],
"defaultPrompt": [
"Create a realistic mock customer dataset.",
"Generate synthetic sales data for Excel.",
"Make reproducible JSON fixtures for my app."
],
"brandColor": "#6C63FF",
"composerIcon": "./assets/faker-composer-icon-32.png",
"logo": "./assets/faker-logo-512.png"
}In the above code, the lines to concentrate on are the final three lines:
composerIcon is the smaller icon shown when selecting or using the plugin.
logo is the larger image used when presenting the plugin.
brandColor gives the plugin a matching accent colour.
Image sizes should be 512x512 pixels for the logo and 32x32 pixels for the composer icon.
Keep the paths relative to the plugin's main folder and begin them with ./
"./assets/faker-composer-icon-32.png"
"./assets/faker-logo-512.png"
After saving the changes, restart the ChatGPT desktop app. If the old icon remains, refresh or reinstall the local plugin so that its cached copy is updated.
Here is what your plugins folder should look like:
Note the addition of an assets folder. You should create an assets folder for yourself here. Then place your two images inside of your assets folder.
Now exit Codex. Restart and, fingers crossed, you should see your new icon in place of the default one:
And here is the main page:
Notice the three smaller icons to the left of Faker Mock Data in the middle.
If you want to use Faker in the future to create some data for a project, you can click the + symbol in the prompt box to see a menu appear:
You will then see your Plugin listed. Note the new icon. Much better than that old, default one we had!
But that's it for this lesson. You now know not only what a Codex plugin is, but also how to create them.
In the next lesson, you'll learn about MCP servers and why they are so big in the Agentic AI world. You'll get to hook up to Blender to create 3D models - all by typing into a prompt window!