GitHub Copilot's verified MCP connectors: connect a service once, use it across the app, CLI, VS Code and cloud

GitHub is rolling out GitHub-verified MCP servers, which it calls connectors, in public preview. Its changelog from 28 October puts the change simply: “With them, you only have to connect once.”

Until now, giving Copilot context from outside your repository meant finding a Model Context Protocol (MCP) server, setting up sign-in, and repeating that in every tool. With connectors, GitHub says, “There’s no server to find, nothing to install locally, and no credentials to manage.”

What you get

• One connection, everywhere. A service connected in one place works in the GitHub Copilot app, Copilot CLI, VS Code, Copilot in the cloud and GitHub Mobile.
• 26 services at launch, including Slack, Figma, Notion, Sentry, Linear, Atlassian Rovo, Cloudflare, Vercel and Work IQ.
• Used when relevant. Copilot calls a connected service’s tools when they fit your request. In the Copilot app you can also schedule automations that use a connected service.

How to connect

• Copilot app: open Customize and select Connect on the service.
• Copilot CLI: run /mcp, switch to the Connectors tab and select Connect.
• VS Code: open the Agents window, go to Customizations and connect the service.
• github.com: browse services on the Copilot customization page.

For Business and Enterprise

The preview is open to all Copilot plans, but on Copilot Business or Enterprise an admin must turn on the Editor preview features policy. If your enterprise uses an MCP allowlist in managed settings, an admin also has to add GitHub-verified MCP servers to it. Some services need their own admin switch too, such as allowing Slack MCP access.

What is not described

GitHub says “Copilot keeps you signed in in the background.” The changelog does not say how those tokens are stored, how long they last, or what scopes each connector asks for. Check each service’s consent screen when you connect.

This builds on recent Copilot agent features such as dynamic workflows and assisted permissions, which let a model approve commands it judges safe.

Why this matters

Teams that maintain hand-configured MCP servers per editor can drop much of that setup, and admins now have one place to decide which services Copilot can reach. Before enabling the policy, review which of the 26 services your organization actually wants agents to read, since one connection now follows a user into the CLI and cloud sessions too. The old rule still holds: a token for one MCP server belongs to that server, so grant the narrowest scope each consent screen allows.