ChatGPT Plugin Examples: What OpenAI Documents and What to Learn From Each

Lists of "top ChatGPT plugins" are easy to write and hard to trust. Most are compiled from memory or from another list.

This one only includes plugins that OpenAI itself names in its documentation or announcements, as read on 9 October 2026. We have not reviewed how well each works, and we are not ranking them. The point is what each shows about design.

Examples OpenAI documents

Canva: a home in the sidebar

OpenAI's extensions page uses Canva as its example of sidebar tabs. A sidebar entry opens the plugin's own interface full screen inside ChatGPT.

What it shows. A sidebar home suits products people open to browse and create, where there is a workspace to return to. Most business software does not need one on day one.

Figma: mentioning items in the composer

Figma is the documented example of composer mentions, where a person types to find and select a specific item from the plugin, such as a file or a frame, and puts it into their prompt.

What it shows. Mentions solve the "which one?" problem. If your product has many similar objects, letting people point at one beats asking the model to guess. OpenAI's page notes that composer mentions work only in the desktop app for now.

Adobe: opening file types

Adobe is the documented example of file extension handlers, for Acrobat and Photoshop. The plugin registers file types and ChatGPT opens those files in the plugin's viewer.

What it shows. If your product owns a file format, a viewer inside the conversation is a natural fit. If it does not, skip this.

Notion and Figma in the plugins repository

OpenAI's packaging guide links to example plugin packages in its public openai/plugins repository, including Notion and Figma.

What it shows. These are worth reading for structure: how a manifest, skills and an MCP reference are laid out in a real package.

Meetings: OpenAI's own plugin

Announced at DevDay 2026, the Meetings plugin takes notes and saves summaries with action items. OpenAI lists it as a beta in the desktop app on macOS for Pro and Business users.

What it shows. OpenAI builds first party features as plugins on the same platform it opened to developers. That is a signal about how seriously it treats the surface, and a reminder that OpenAI may ship a plugin in your category.

Airtable

OpenAI hosts a page for the Airtable plugin on its business site. It describes asking questions, exploring roadmaps and updating records across Airtable bases from ChatGPT. Under "What's included" it lists one app, Airtable, and three skills.

What it shows. Two things. Structured records in a customer's own account are the classic plugin case: questions and updates over data ChatGPT could never reach alone. And the package is a real mix: one connection to the product plus written skills that teach ChatGPT how to use it well.

Bits and Bolts: the reference plugin

Bits and Bolts is OpenAI's sample plugin in the openai/mcp-extensions repository. It exists to demonstrate the extension surfaces with working code.

What it shows. If you plan to build any interface extension, read this code first.

The project board automation

OpenAI's DevDay recap gives one example of an event driven plugin: you ask ChatGPT to watch a project board, and when a task arrives it reads the linked documents and drafts a plan while you are away.

What it shows. Events fit workflows that begin with "when this happens". The underlying specification is still a draft.

What all of these have in common

Read the list again and three things stand out.

1. Each holds something ChatGPT cannot reach alone: designs, documents, records, meetings.
2. Each is used repeatedly by the same person.
3. Each already had an API and accounts before any plugin existed.

That is the profile of a strong plugin candidate. It is also what our readiness checker tests for.

Patterns a typical SaaS can copy

You are probably not Canva. These five patterns fit ordinary business software and need no interface extensions at all.

1. Ask my account. Read tools over the customer's own data. "Which invoices are overdue?" "What did we ship to this customer last month?" Lowest risk, fastest through review.

2. Do the routine thing. One or two write tools for the action people repeat most: create the ticket, book the slot, log the call. ChatGPT asks for confirmation where it matters.

3. Prepare, then hand over. The plugin gathers what is needed and returns a link that opens your product at the right place to finish. Useful when the last step is too sensitive or too visual for a conversation.

4. Explain my data. Return structured records and let ChatGPT do the summarising, comparing and charting. You supply facts. It supplies analysis.

5. Combine with everything else. Your data plus the person's files, notes and other plugins in one request. This is the thing your own dashboard cannot do.

Patterns to avoid

The brochure. A plugin that only describes your product. OpenAI does not accept trial or demo plugins, and nobody installs an advert.

The whole API. Forty tools mirroring forty endpoints. The model cannot choose between them reliably and you cannot test them all.

The checkout. A plugin designed to sell a subscription. OpenAI's guidelines do not allow digital products or subscriptions to be sold through plugins.

Building yours

Start from the questions your customers would ask, not from your endpoint list. Three to five tools is a good first release.

• See the full process and prices on the ChatGPT plugin development page: from $4,999 fixed, submitted for OpenAI review in 14 days
• Understand the server underneath in MCP for ChatGPT plugins
• Know what review will ask for in the submission guide

HouseofMVPs is an independent studio. None of the companies named here are clients, and we are not affiliated with OpenAI.