How to Add MCP Servers to VS Code (Copilot Chat Setup)

How MCP Works in VS Code

VS Code supports MCP servers through GitHub Copilot Chat. When you configure an MCP server, Copilot Chat gains access to external tools. Ask it to query your database, check deployment status, or pull metrics from any connected service.

MCP tools appear in the Copilot Chat tool picker. The agent decides when to call them based on your prompt context.

Step 1: Create .vscode/mcp.json

VS Code reads MCP configuration from .vscode/mcp.json in your workspace root:

{
  "servers": {
    "your-tool": {
      "type": "sse",
      "url": "https://datafaucet.dev/api/mcp/YOUR_SERVER_ID/sse"
    }
  }
}

Replace YOUR_SERVER_ID with your DataFaucet server ID from your dashboard.

Step 2: Enable MCP in Settings

Open VS Code Settings (Ctrl+Comma) and search for "mcp". Ensure chat.mcp.enabled is set to true. This is enabled by default in recent VS Code versions but worth confirming.

Step 3: Verify the Connection

Open the Command Palette (Ctrl+Shift+P) and run "MCP: List Servers". Your server should appear with a connected status. If disconnected, check the URL and ensure Copilot Chat extension is installed and active.

Step 4: Use Tools in Copilot Chat

Open Copilot Chat (Ctrl+Shift+I) and switch to Agent mode. Type a prompt that requires your MCP tool:

• "Check the latest orders from our API"
• "What's the current error rate in production?"
• "Show open tickets assigned to me"

Copilot identifies the relevant tool, calls it, and weaves the result into its response.

Adding Multiple Servers

Stack multiple servers in the same config:

{
  "servers": {
    "database": {
      "type": "sse",
      "url": "https://datafaucet.dev/api/mcp/SERVER_1/sse"
    },
    "monitoring": {
      "type": "sse",
      "url": "https://datafaucet.dev/api/mcp/SERVER_2/sse"
    },
    "tickets": {
      "type": "sse",
      "url": "https://datafaucet.dev/api/mcp/SERVER_3/sse"
    }
  }
}

Each server name becomes a tool namespace. Copilot sees all available tools and selects the appropriate one.

SSE vs Stdio Transports

VS Code supports both:

• SSE (Server-Sent Events): Remote hosted servers. DataFaucet uses this. Provide the URL, nothing to install locally.
• Stdio: Local processes VS Code spawns. Requires specifying a command and arguments.

{
  "servers": {
    "local-tool": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@some/mcp-server"]
    }
  }
}

DataFaucet servers are always SSE. No local dependencies, no Docker, no process management.

User vs Workspace Config

You can also add MCP servers to your user-level settings.json:

{
  "mcp": {
    "servers": {
      "your-tool": {
        "type": "sse",
        "url": "https://datafaucet.dev/api/mcp/YOUR_SERVER_ID/sse"
      }
    }
  }
}

User-level servers are available across all workspaces. Workspace-level (.vscode/mcp.json) servers only load for that project. Use workspace config for project-specific tools and user settings for tools you want everywhere.

Troubleshooting

Server shows disconnected:

Confirm the SSE URL is correct. Visit it in a browser to verify the endpoint responds. Check that GitHub Copilot Chat extension is active (not just GitHub Copilot).

Tools not appearing:

Ensure you're in Agent mode within Copilot Chat. Standard chat mode doesn't invoke MCP tools. Also verify chat.mcp.enabled is true in settings.

Auth errors on tool calls:

If the target API needs fresh credentials, re-record the server on your DataFaucet dashboard. Browse the app again to capture updated auth tokens.

Creating Your First Server

No MCP server yet? DataFaucet creates one in 60 seconds:

1. Paste the URL of any web app on datafaucet.dev
2. Browse the app in the cloud browser for 60 seconds
3. DataFaucet captures API calls and deploys a hosted MCP server
4. Copy the SSE URL into your .vscode/mcp.json

Works with internal tools, undocumented APIs, and apps that have no official MCP integration. Your AI gets typed tool definitions, pre-configured auth, and structured responses.