Quickstart

Plugins extend and customize ChatGPT and Codex. They can add capabilities, connect to external services, or both. A plugin can include skills that provide instructions and resources, an MCP server that exposes tools, or both.

ChatGPT and Codex share one universal plugin directory. Public plugins are published once and become discoverable from supported surfaces in both products.

This tutorial creates a small skills-only plugin. It is the fastest way to learn the plugin package structure before you add an MCP server or custom UI.

By the end, you will have a plugin that teaches ChatGPT and Codex how to turn meeting notes into a consistent follow-up.

Create the plugin

The fastest way to start is to ask the built-in plugin creator:

@plugin-creator Create a plugin named meeting-follow-up.
Add a skill that turns meeting notes into a concise recap with decisions,
action items, owners, and a customer follow-up draft.

In Codex, invoke the same skill as $plugin-creator.

The generated plugin should have this structure:

Review the manifest

The plugin manifest is stored at .codex-plugin/plugin.json. For this skills-only plugin, it points to the bundled skills directory:

{
  "name": "meeting-follow-up",
  "version": "1.0.0",
  "description": "Turn meeting notes into consistent follow-ups",
  "skills": "./skills/"
}

Use a stable, kebab-case name. Keep the description focused on the outcome the plugin helps someone achieve.

Review the skill

Each skill lives in its own directory and includes a SKILL.md file. A minimal version looks like this:

---
name: meeting-follow-up
description: Turn meeting notes into a recap, action register, and customer follow-up draft.
---

Use this skill when the user provides meeting notes or a transcript and asks
for a follow-up.

1. Summarize the outcome in two or three sentences.
2. List decisions that were made.
3. List action items with an owner when the source identifies one.
4. Draft a concise customer follow-up email.

Do not invent owners, deadlines, decisions, or commitments that are missing
from the source.

Adjust the instructions to match the workflow you want to make repeatable. Keep inputs, required sections, and boundaries explicit.

Test the plugin

Ask the plugin creator to add the plugin to a personal marketplace for local testing. Restart the ChatGPT desktop app, open the Plugins Directory, select your local source, and install the plugin. This personal marketplace is a local testing source; it does not publish the plugin to the universal directory.

Start a new chat with the plugin enabled and provide representative meeting notes:

Use the meeting follow-up skill on these notes. Keep unknown owners and dates
unassigned.

We agreed to move the launch to the first week of August. Priya will update the
rollout checklist. The customer wants a revised security summary before the
next review.

Verify that the response includes the decision, assigns only the known action owner, and does not invent an owner or date for the security summary.

Add an MCP server when you need tools

This plugin does not need an MCP server because its workflow can use information already provided to the model. Add an MCP server when the plugin needs to connect to a service, authenticate users, expose controlled tools, or run code on infrastructure you operate.

Continue with the MCP server and UI quickstart, or read Plugin architecture to compare plugin shapes. Custom UI remains optional even when a plugin contains an MCP server.

Publish the plugin

When the plugin is ready for other people, review the complete plugin build guide. To publish it publicly, use the plugin submission portal.