Code Assist MCP for Klaviyo

Build 25.0.9540

Using With GitHub Copilot

GitHub Copilot is an AI-powered assistant that supports natural language interaction in Visual Studio Code. With CData's Model Context Protocol (MCP), Copilot can securely query and describe live data from CData Code Assist MCP using plain language prompts.

This topic explains how to configure the CData Code Assist MCP to work with GitHub Copilot in Visual Studio Code.

Prerequisites

Before you begin, make sure the following are installed and configured:

  • The CData Code Assist MCP for Klaviyo is installed.
  • A connection has been created and tested in the CData Code Assist MCP Configuration Tool.
  • Visual Studio Code is installed on your machine.
  • GitHub Copilot Chat extension is enabled in Visual Studio Code. For more details, see GitHub Copilot Chat on Visual Studio Marketplace.

Using Claude Desktop Configuration

If a connection has been configured for Claude Desktop, GitHub Copilot can discover and use that configuration.

To enable GitHub Copilot to use CData Code Assist MCP servers configured in Claude Desktop:

  1. In Visual Studio Code, select File > Preferences > Settings.
  2. Search for "MCP".
  3. Enable Claude Desktop configuration by selecting the checkbox.
  4. Open GitHub Copilot Chat and select the tool icon.
  5. Enable the MCP server you want to use by selecting the checkbox next to its name.

This method is the simplest way to use an existing MCP server and does not require editing configuration files manually.

Configuring CData Code Assist MCP

GitHub Copilot loads MCP server definitions from mcp.json files only. To register an MCP server with GitHub Copilot Chat, create a file named mcp.json and reference it inside your Copilot Chat configuration in the appropriate location. The location of this file determines the scope.

The recommended way to populate mcp.json is to copy the MCP server JSON configuration directly from the CData Code Assist MCP Configuration Tool UI and paste it into the appropriate file. This ensures all paths, arguments, and connection names are accurate and eliminates the need to edit placeholders manually.

Configuration Scopes

GitHub Copilot in Visual Studio Code supports three types of configuration scopes for MCP servers:

ScopeLocationShareableDescription
From Claude Desktop%APPDATA%\Claude\claude_desktop_config.jsonNoUses the MCP server configuration already created for Claude Desktop. GitHub Copilot can discover this configuration automatically.
User%APPDATA%\Roaming\Code\User\mcp.json on WindowsNoApplies to all Visual Studio Code workspaces on the machine. This configuration is personal and not intended for sharing across machines or repositories.
Project<workspace-root>/.vscode/mcp.json inside the project folder
For example, C:\Projects\MyProject\.vscode\mcp.json on Windows		YesApplies only to the current workspace. This configuration can be version-controlled and shared with team members.

Steps

  1. Create a file named mcp.json in the appropriate location based on your desired configuration scope.
  2. Copy the MCP server JSON configuration from the CData Code Assist MCP Configuration Tool UI.
  3. Paste the configuration into the mcp.json file.
  4. Verify that the server name matches the MCP connection name defined in the CData Code Assist MCP Configuration Tool.
  5. Ensure the GitHub Copilot Chat extension is installed and enabled.
  6. Restart Visual Studio Code so GitHub Copilot can load the updated MCP configuration.
  7. Open GitHub Copilot Chat and use the tool selector to enable the configured MCP server.

Example Configuration

The following example illustrates the structure of an mcp.json entry and explains how key values are used by GitHub Copilot.

  • Connection name: The server name (for example, mydatasource_sbx) must match the connection name defined in the CData Code Assist MCP Configuration Tool exactly and is case-sensitive.
  • Java path: The command value specifies the path to the Java executable bundled with the MCP server.
  • JAR path: The args array includes the path to the MCP server JAR file and the connection name passed at startup.

{
  "servers": {
    "mydatasource_sbx": {
      "command": "C:\\Program Files\\CData\\CData Code Assist MCP for Klaviyo 2025\\jre\\bin\\java.exe",
      "args": [
        "-Dfile.encoding=UTF-8",
        "-jar",
        "C:/Program Files/CData/CData Code Assist MCP for Klaviyo 2025/lib/cdata.mcp.klaviyo.jar",
        "mydatasource_sbx"
      ]
    }
  }
}

Verifying Configuration

After you've added your server configuration, restart Visual Studio Code and verify that the MCP server is recognized.

Steps

  1. Restart Visual Studio Code.
  2. Open the Command Palette using Ctrl+Shift+P on Windows.
  3. Search for GitHub Copilot: Open Chat and launch the Copilot Chat interface.
  4. Enter a test query using your MCP server name.

Sample Prompts

  • Show me the available tables in mydatasource_sbx
  • Query the Account table and show me the first 5 records
  • List fields in the Contact object

Managing Multiple MCP Servers

You can configure multiple MCP servers by including them in the same configuration block. This is useful if you work with different environments such as staging and production or connect to multiple data sources.

Each entry must use the exact connection name as configured in the CData Code Assist MCP Configuration Tool. This name is case-sensitive and is passed as the final argument in the args array. 

{
  "servers": {
    "mydatasource_sbx": {
      "command": "...",
      "args": ["...", "mydatasource_sbx"]
    },
    "mydatasource_prod": {
      "command": "...",
      "args": ["...", "mydatasource_prod"]
    }
  }
}

Troubleshooting

The server is not recognized

  • Restart Visual Studio Code after saving your settings file. GitHub Copilot loads MCP server definitions only when Visual Studio Code starts, so any changes to mcp.json will not take effect until you relaunch the editor.
  • Check for typos or case mismatches in the server name. The name you provide in the configuration must match the MCP connection name exactly, including capitalization. If the names do not match, Copilot will not associate your prompts with the correct MCP server.
  • Ensure the paths to Java and the JAR file are correct. If either path is incorrect or the files are missing, the MCP server cannot start, and Copilot cannot connect to it.
  • Validate your JSON using a free online tool. A single missing comma or bracket can prevent Visual Studio Code from loading the configuration entirely, which causes Copilot to ignore the MCP server block.

File path issues

  • Use either double backslashes (C:\\Program Files\\...) or forward slashes (C:/Program Files/...) consistently across both the command and args settings. Both formats are supported, but consistency avoids errors and improves readability.
  • If the MCP server does not start, verify that the folders and filenames in your path match your installed MCP server for Klaviyo. Some sources have different naming conventions, and version years must match your installed driver.

Authentication errors

  • Open the CData Code Assist MCP Configuration Tool to confirm that your connection is still valid. Copilot relies on the same authentication settings used by the MCP server, so incorrect or expired credentials can cause failures.
  • Select your configured connection and click Save & Test to identify whether the issue is related to credentials, network access, or the underlying service.
  • Update your credentials if needed, and save the connection. Copilot cannot authenticate through Visual Studio Code, so all credentials must be stored and validated through the CData Code Assist MCP Configuration Tool.
  • Restart Visual Studio Code after updating your MCP server configuration. Copilot must reload the MCP server process to pick up any new authentication settings.

The MCP server starts and stops immediately

  • This usually indicates an issue with the Java or JAR path. Open a PowerShell window and manually run the Java command that launches your MCP server to confirm the paths are correct.
  • If running the server manually shows an error, verify that your JAR file is in the correct folder and that the version of Java bundled with your MCP server has not been removed or modified.

Copilot does not return data

  • Confirm that Copilot is calling the correct MCP server. Use queries that reference the server name directly to avoid ambiguity, especially when multiple MCP servers are configured.
  • Verify that your connection type supports the operation you are requesting. For example, some APIs may require elevated permissions or may not expose certain metadata.
  • Open the Output panel in Visual Studio Code and check for MCP-related logs. Errors such as permission denied, invalid API keys, or network connection failures may appear there.

Copyright (c) 2026 CData Software, Inc. - All rights reserved.
Build 25.0.9540