What Are ChatGPT Plugins? How They Differ from MCP & Skills
Still drafting in ChatGPT, then copying the results into your design tool, your docs, and your team chat one by one? Every hop between tools breaks the context, and you end up explaining the same thing over and over.
A ChatGPT plugin is an installable bundle that packages skills (instructions for how to do a job) and MCP servers (connections to outside services) so ChatGPT can work inside the tools you already use. ChatGPT and Codex share a single plugin directory. The daglo team put this guide together from what we heard at OpenAI DevDay Seoul 2026 and from OpenAI's official documentation.
What is a ChatGPT plugin?
A ChatGPT plugin is a bundle you install once that can contain skills, an MCP server, or both. Skills teach ChatGPT how to do a task; MCP servers give it access to data and actions in outside systems.
According to OpenAI's plugin quickstart, plugins can add capabilities, connect to external services, or both. A public plugin is published once and can be found from both ChatGPT and Codex.
One note on naming: these are not the ChatGPT plugins you may remember from 2023. That earlier beta stopped accepting new use in 2024, and GPTs took over its role. Today's plugins are a new structure built around skills and MCP.
Plugins vs. MCP vs. skills vs. GPTs
Think of MCP as the pipe to outside tools, a skill as the playbook, and a plugin as the package that ships both. GPTs are custom versions of ChatGPT built from instructions and files.
| Type | What it is | What it does |
|---|---|---|
| MCP server | A standard connection between AI and an outside system | Runs tools such as looking up data or editing a document |
| Skill | Instructions and reference material for a specific workflow | Tells the AI how a product works and what steps to follow |
| Plugin | An installable bundle of skills and MCP servers | Found in the directory and installed in one step |
| GPTs | A custom ChatGPT with its own instructions and knowledge files | Acts as a chat partner with a fixed role |
The practical difference: an MCP server on its own only gives ChatGPT tools, and a skill on its own only gives it know-how. A plugin delivers both together, so a service can hand over its tools and the right way to use them in one install. OpenAI also publishes a guide for moving existing custom GPT workflows over to plugins. If you want to connect MCP servers yourself, see our MCP guide.
What plugins can do: new surfaces and event-triggered work
Plugins aren't limited to the chat window. They can open their own interface in the sidebar or beside a conversation, and they can start AI work when something changes in a connected app.
| Extension | What it does |
|---|---|
| Sidebar apps | Open a service from the sidebar and work in it full screen |
| Conversation panels | Open a service beside the chat so work and conversation stay together |
| Composer mentions | Find and pick plugin content from the message box (desktop app) |
| File viewers and editors | Open a service's own file types in its own interface and save changes |
| Onboarding and settings | Guide setup and manage product-specific settings inside ChatGPT |
| MCP Events | Start work when a new message, comment, or status change arrives |
At the Seoul event, presenters showed a design tool example: you select an element on screen, it's passed to ChatGPT, and the edited result shows up back in the interface. OpenAI's MCP Events documentation explains the event side: you tell ChatGPT what to watch and how to respond, and when an update arrives, ChatGPT handles it in that chat following your instructions. The docs give examples like watching a channel for bug reports or a document for review comments, and the Seoul talk mentioned triggers like a new pull request or a meeting status change.
Discovery is changing too. ChatGPT can suggest a plugin during a conversation when it fits your request. In Seoul, presenters said recommendations take signals such as how often installs turn into real use and ongoing activity into account, and cited roughly 4,000 plugins in the directory (figures as presented at the event).
How marketers can use plugins, and what to check first
For marketers and planners, plugins cut down the back-and-forth between tools. But connecting one gives AI access to company data, so check permissions and scope before you install.
- Doc reviews: When someone comments on a campaign brief, have ChatGPT draft the revision.
- Creative tweaks: Select a banner headline in your design tool and rewrite it in the chat.
- Inbox triage: When new messages land in a shared channel, have them summarized and sorted.
Before you connect, keep a few things in mind. Installing a plugin doesn't open up data access by itself: each app still needs its own sign-in, and the connected service's permissions still apply. Check that the permissions requested match what the job actually needs, decide when ChatGPT should ask before acting, and confirm your workspace admin allows it. For a similar setup in another tool, see our guide to Claude Code plugins.
Meeting records become fuel for AI work
As plugins and events spread, the trigger for AI work increasingly comes from data that carries business context. Meeting records, with what was decided and who owns what, are a prime example.
daglo transcribes and summarizes your meetings, keeping the full transcript and the summary side by side. With Board Chat, you can ask something like "List only the dates we agreed on in this meeting" and get an answer grounded in the record. Speaker ID remembers the voices of the people in your meetings and fills in their real names instead of "Speaker 1" from the next recording on. daglo is the first service in Korea to offer this to all users. Learn more about Speaker ID
For the rest of the announcements, read our OpenAI DevDay Seoul 2026 recap.
#ChatGPTPlugins #ChatGPT #MCP #AISkills #PluginDirectory #AIAgents #SpeakerID #daglo
Where talk becomes work.
Get started with 1,000 credits a month — no credit card required.