Pular para o conteúdo principal

MCP (Model Control Protocol)

CodeGPT now supports the integration of external tools via the Model Context Protocol (MCP), allowing AI models to interact with various services and data sources. This tutorial will walk you through the process of configuring an MCP server within CodeGPT.

CodeGPT MCP Host

Understanding MCP

What is an MCP Host?

An MCP Host is the client application that connects to MCP servers, facilitating communication between AI models and external tools. In this context, CodeGPT acts as the MCP Host, managing connections and interactions with configured MCP servers.​

What is an MCP Server?

An MCP Server is a service that exposes tools and functionalities to the MCP Host. These tools can perform various tasks, such as file operations, database queries, or API interactions. MCP servers can be run locally or remotely and communicate with the host using standardized protocols.​

Configuring an MCP Server in CodeGPT

Follow these steps to add and configure an MCP server in CodeGPT:

  1. Open Settings Click on the gear icon (⚙️) located in the sidebar to open the Settings menu.​
Screenshot 2025-04-28 at 12 20 35
  1. Navigate to the MCP Configuration Tab In the Settings menu, select the MCP Configuration tab to access MCP-related settings.​

  2. Open the MCP Config File Click on Open MCP Config File to open the configuration file (mcp.json) where you can define your MCP servers.​

Screenshot 2025-04-28 at 12 21 08
  1. Add Your Server Configuration In the mcp.json file, add your server configuration. For example, to add a filesystem server:​
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-filesystem",
"/Users/username/Desktop",
"/path/to/other/allowed/dir"
]
}
}
}
  • "command" specifies the executable to run (e.g., npx).
  • "args" provides the arguments for the command, including the server package and any necessary parameters.​
  1. Refresh Server Connections After saving your configuration, return to the MCP Configuration tab and click on Refresh Server Connections to apply the changes and start the new server.​

Exploring Available Tools

Once the server is running:​

  • You'll see a list of available servers.

  • Click on a server to view its exposed tools and their descriptions.

  • Select the tools you wish to enable for use by the AI models.​

Screenshot 2025-04-28 at 12 22 08

This setup allows you to customize the capabilities available to your models, enhancing their functionality based on your specific needs.