OpenAI MCP Extensions (GitHub Repo)
GitHub ● Covered by 7 sources
OpenAI put out MCP extensions for ChatGPT-style plugins. They add sidebar tools, file viewers, search and custom forms that feel built in.
Based on reporting by GitHub — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI’s MCP Extensions project is aimed at one thing: making plugins for ChatGPT behave less like add-ons and more like part of the product. The GitHub repo shows how developers can build features that show up in places users already know, rather than forcing everything into a separate panel or awkward workaround.
The examples are built around a plugin called Bits & Bolts, which serves as the demo for the extension types. One path is sidebar entrypoints, letting an app appear right in the sidebar. Another is file extension handlers, which can render a custom viewer when a supported file type is opened. There’s also composer mentions, so users can search a plugin’s resources from the composer and drop references into a message.
And then there are extended forms, which push the idea further. The example in the repo lets someone choose a CAD part using thumbnail options. That’s the kind of detail that tells you what this project is really for: not generic plugin plumbing, but tighter, more native-feeling interactions inside ChatGPT.
Getting started is straightforward on paper. Install the Bits & Bolts Remote plugin from the plugin directory, then select it in the sidebar to open the Parts Library. After that, developers are told to create a plugin, add the SDKs to their MCP server, and read the spec for the supported extensions. The SDK names are different depending on the stack: @openai/mcp-extensions for TypeScript and openai-mcp-extensions for Python. The project is licensed under Apache 2.0.
My take — AI-written commentary, not fact-checked reporting
This is the sensible version of AI platform expansion: fewer flashy promises, more places where the tool actually lives in the workflow. The real tell is that OpenAI is pushing sidebar entries, file handlers, and composer hooks instead of another grand “ecosystem” speech. That usually means the boring part is getting serious, which is where products either become useful or stay a demo forever.
Read more about this at: GitHub
