OpenAI brings ChatGPT extensions to the sidebar via MCP

Third-party software breaks out of the chat stream into dedicated panels, full-window views, and desktop shortcuts built on an open standard.

OpenAI is fundamentally changing how people interact with its flagship tool by introducing ChatGPT extensions built on the Model Context Protocol (MCP). In a technical demonstration shared on X by the OpenAI Developers team on October 7, 2026, external integrations no longer appeared solely as blocks of generated text inside a conversation thread. Instead, applications now occupy dedicated user interface areas, including persistent full-width sidebar views and contextual side panels running alongside active chats.

How ChatGPT extensions change daily workflows

Until now, calling a plugin or custom GPT meant watching text scroll sequentially down a single conversation window. That format created friction for visual design work or complex document reviews. The new architecture distributes external tools across four distinct interface points:

Global fullscreen views: opened directly from the left navigation sidebar for focused application work.
Thread side panels: contextual secondary windows that sit alongside an active discussion without breaking conversation history.
Direct file handlers: integrated viewers and editors that open attachments right inside the workspace.
Configuration pages: dedicated settings panels for authentication, permissions, and service controls.

Early partner rollouts show how this layout functions in practice. Canva runs inside the sidebar as an interactive visual design board. Figma hooks into the desktop composer through @ mentions, allowing users to locate and attach live design files without leaving their chat window. Meanwhile, Adobe added custom handlers that open Acrobat PDFs and Photoshop assets directly in the browser interface.

Technical standardization with the open MCP standard

The central technical choice behind the launch is OpenAI’s adoption of the Model Context Protocol (MCP), an open specification originally developed by Anthropic to standardize data transfer between language models and outside tools. Instead of forcing third-party teams to build custom visual bridges from scratch, OpenAI rolled out the official @openai/mcp-extensions package with software development kits (SDKs) for TypeScript and Python.

The architecture handles standard tool execution while introducing MCP Events. This addition lets third-party services push notifications and trigger interactive flows directly into an active ChatGPT session. To speed up developer onboarding, OpenAI published “Bits & Bolts”, an open-source reference plugin that provides boilerplate code for every supported surface area.

Availability, platform limits, and unconfirmed claims

Web access to the new extensions is initially reserved for paid accounts across Plus, Pro, Business, and Enterprise tiers. Free users and basic Go tier subscribers will receive web support at a later stage, though OpenAI has not committed to a firm timeline. Any reports citing specific release dates for free accounts or immediate plans for a paid commercial marketplace are currently RUMOR, as official documentation only notes that expanded access is coming soon.

Initial platform restrictions also remain in place. Composer @ mentions are exclusive to the ChatGPT desktop client on macOS and Windows, excluding web browsers and mobile apps on iOS and Android for now. By moving connectors out of the chat feed into permanent interface real estate, OpenAI is positioning ChatGPT less as a simple question-and-answer bot and more as an everyday operational workspace.