Extending Copilot Chat with the Model Context Protocol (MCP)

Overview

The Model Context Protocol (MCP) is an open standard that defines how applications share context with large language models (LLMs). MCP provides a standardized way to connect AI models to different data sources and tools, enabling them to work together more effectively.

You can use MCP to extend the capabilities of Copilot Chat by integrating it with a wide range of existing tools and services. For example, the GitHub MCP server allows you to use Copilot Chat in Visual Studio Code to perform tasks on GitHub. You can also use MCP to create new tools and services that work with Copilot Chat, allowing you to customize and enhance your experience.

For more information on MCP, see the official MCP documentation.

For information on some of the other currently available MCP servers, see the MCP servers repository.

Prerequisites

• Access to Copilot. 有关如何访问 Copilot 的信息，请参阅“什么是 GitHub Copilot？”。
• Visual Studio Code version 1.99 or later. For information on installing Visual Studio Code, see the Visual Studio Code download page.

Configuring MCP servers in Visual Studio Code

To configure MCP servers in Visual Studio Code, you need to set up a configuration script that specifies the details of the MCP servers you want to use. You can configure MCP servers for either:

• A specific repository. This enables you to share MCP servers with anyone who opens the project in Visual Studio Code. To do this, create a .vscode/mcp.json file in the root of your repository.

• Your personal instance of Visual Studio Code. You will be the only person who has access to configured MCP servers. To do this, add the configuration to your settings.json file in Visual Studio Code. MCP servers configured this way will be available in all workspaces.

  注意

  We recommend you use only one location per server. Adding the same server to both locations may cause conflicts and unexpected behavior.

The steps below show how to configure the Fetch MCP server in your .vscode/mcp.json file. The Fetch MCP server is a simple MCP server that provides web content fetching capabilities. For more information on the Fetch MCP server, see the Fetch directory in the MCP Server repository.

You can use the same steps to configure MCP servers in your personal Visual Studio Code settings. Details on how to configure other MCP servers are available in the MCP servers repository.

For information on configuring the GitHub MCP server, see Using the GitHub MCP Server.

1. Add the following configuration to your .vscode/mcp.json file:

   JSON

   {
   "inputs": [
     // The "inputs" section defines the inputs required for the MCP server configuration.
     {
       "type": "promptString"
     }
   ],
   "servers": {
     // The "servers" section defines the MCP servers you want to use.
     "fetch": {
       "command": "uvx",
       "args": ["mcp-server-fetch"]
     }
    }
   }
   

2. Save the .vscode/mcp.json file.

3. A "Start" button will appear in your .vscode/mcp.json file, at the top of the list of servers. Click the "Start" button to start the MCP servers. This will trigger the input dialog and discover the server tools, which are then stored for later sessions.

   

4. Open Copilot Chat by clicking the icon in the title bar of Visual Studio Code.

5. In the Copilot Chat box, select Agent from the popup menu.

   

6. To view your list of available MCP servers, click the tools icon in the top left corner of the chat box. This will open the MCP server list, where you can see all the MCP servers and associated tools that are currently available in your Visual Studio Code instance.

   • Optionally, you can define toolsets, groups of related tools that you can reference in chat. Toolsets make it easier to group related MCP tools together and quickly enable or disable them. For information on how to define and use a toolset, see the VS Code docs.

For more information on configuring MCP servers in Visual Studio Code, see Use MCP servers in Visual Studio Code (Preview) in the Visual Studio Code documentation.

Using MCP servers in Copilot Chat

Once you have configured your MCP servers, you can use them in Copilot Chat to access a wide range of tools and services. In the example below, we will use the Fetch MCP server to fetch details about a web page.

1. 若要打开聊天视图，请单击活动栏中的聊天图标或按 Control+Command+i (Mac) / Ctrl+Alt+i (Windows/Linux)。

2. In the Copilot Chat box, select Agent from the popup menu.

3. In the file with the MCP configuration, check that the MCP server is running. If it is not running, click the "Start" button to start the MCP server.

   

4. Ask Copilot Chat to fetch the details of a URL. For example:

   Fetch https://github.com/github/docs.

5. If Copilot asks you to confirm that you want to proceed, click Continue.

6. Copilot will fetch the details of the URL and display them in the chat box.

Optionally, you can use MCP prompts and resources in VS Code.

• MCP servers can define preconfigured prompts for interacting with their tools. You can access these prompts in chat with slash commands, using the format /mcp.servername.promptname.
• MCP servers provide resources, which are any kind of data that the server wants to make available. For example, the GitHub MCP server provides repository content as a resource. To add resources from an MCP server to your chat context, click Add Context... in the chat box, then click MCP Resources.

For more information on using MCP servers in Visual Studio Code, see Use MCP servers in Visual Studio Code (Preview) in the Visual Studio Code documentation.

Using existing MCP configurations

If you already have an MCP configuration in Claude Desktop, you can use that configuration in Visual Studio Code to access the same MCP servers. To do this, add the following configuration to your settings.json file in Visual Studio Code:

"chat.mcp.discovery.enabled": true

"chat.mcp.discovery.enabled": true

Visual Studio Code will automatically find your existing configuration and use it in your Visual Studio Code instance.