How to Set Up an MCP Server With Claude and Connect It to Google Drive

AI tools are only as powerful as the data they can reach. So when Anthropic built the Model Context Protocol, they gave Claude a way to reach far beyond a chat window. An MCP server lets Claude connect directly to external tools, files, and services. The result is a smarter, more useful assistant that works with your actual workflow.

This guide walks you through setting up an MCP server with Claude Desktop and linking it to Google Drive. By the end, Claude will be able to read, search, and reference your Drive files in real time.

What Is an MCP Server, Exactly?

MCP stands for Model Context Protocol. It is an open standard that Anthropic developed to let AI models communicate with external data sources and tools. Think of it as a bridge between Claude and the software you already use every day.

An MCP server is the piece of software that sits on that bridge. It handles requests from Claude, fetches the relevant data, and returns it in a format the model understands. Without an MCP server, Claude only knows what you type. With one connected, Claude can see your documents, calendars, emails, and more.

The protocol is open source. Developers can build MCP servers for virtually any platform, and many already exist for popular services like Google Drive, Slack, GitHub, and Notion.

What You Will Need Before You Start

Before touching any configuration files, gather a few things. First, download and install Claude Desktop from Anthropic’s website. This is the desktop application that supports MCP connections. The Claude web interface does not currently support custom MCP servers.

Next, make sure you have Node.js installed on your machine. Most MCP servers run on Node, so you will need version 18 or higher. You can check your version by running node --version in your terminal. Finally, confirm you have a Google account and access to Google Drive. You will use Google’s OAuth process to authorize the MCP server.

Step 1: Locate the Claude Desktop Config File

Claude Desktop reads its MCP server configuration from a single JSON file on your computer. The location depends on your operating system. On macOS, navigate to ~/Library/Application Support/Claude/claude_desktop_config.json. On Windows, the path is %APPDATA%\Claude\claude_desktop_config.json.

Open that file in any text editor. If it does not exist yet, create it. The file should contain valid JSON, so start with an empty object {} if the file is new. This file is where you will register every MCP server you want Claude to use.

Step 2: Install the Google Drive MCP Server

Anthropic and its community maintain a set of reference MCP servers in an open GitHub repository. The Google Drive server is one of them. Open your terminal and run the following command to install it globally via npm:

npm install -g @modelcontextprotocol/server-gdrive

This installs the server package on your machine and makes it available as a command. Once the installation finishes, confirm it worked by running mcp-server-gdrive --version. You should see a version number printed back. If you get an error, double-check your Node.js version and that npm installed cleanly.

Step 3: Create Google OAuth Credentials

The Google Drive MCP server uses OAuth 2.0 to access your files securely. You need to create credentials in the Google Cloud Console before the server can authenticate. Head to console.cloud.google.com and sign in with your Google account.

Create a new project or select an existing one. Then navigate to APIs and Services, click Enable APIs, and search for the Google Drive API. Enable it. After that, go to Credentials, click Create Credentials, and choose OAuth client ID. Select Desktop app as the application type and give it a name. Download the resulting JSON credentials file and save it somewhere easy to find, like your home directory. Rename it gdrive-credentials.json for clarity.

Step 4: Add the Server to Your Config File

Now open your claude_desktop_config.json file again. You will add an entry under a mcpServers key. Here is what the configuration looks like:

{
  "mcpServers": {
    "gdrive": {
      "command": "mcp-server-gdrive",
      "args": [],
      "env": {
        "GDRIVE_CREDENTIALS_PATH": "/path/to/gdrive-credentials.json"
      }
    }
  }
}

Replace /path/to/gdrive-credentials.json with the actual path to the file you downloaded. On macOS this might be /Users/yourname/gdrive-credentials.json. Save the config file when you are done. Each entry in mcpServers is a key-value pair. The key is a label you choose, and the value defines how Claude should launch the server.

Step 5: Authorize Claude to Access Your Drive

Quit Claude Desktop completely and relaunch it. On startup, Claude will read your config file and attempt to connect to the Google Drive MCP server. A browser window should open automatically, prompting you to sign in to Google and grant access to your Drive files.

Complete the authorization flow. Google will display the permissions the app is requesting, which typically include reading file metadata and content. Accept them. The server will store an access token locally so you do not need to re-authorize on every launch. If the browser window does not open, check your terminal for error messages. Common issues include a wrong file path in the config or a Node version that is too old.

Step 6: Test the Connection Inside Claude

Once authorization is complete, go back to Claude Desktop and start a new conversation. Ask Claude something like: “Can you list the most recent files in my Google Drive?” or “Search my Drive for documents about Q3 planning.” Claude should respond with real data pulled from your Drive rather than a generic reply.

You will also notice a small indicator in the Claude interface showing which MCP servers are active. If you do not see Drive data, open Settings in Claude Desktop and check the MCP section. It will show you which servers are connected and whether any errors occurred during startup.

Troubleshooting Common MCP Server Issues

A few problems come up regularly with MCP server setups. The most common is a path error in the config file. JSON does not allow backslashes without escaping, so Windows users need to write paths like C:\\Users\\name\\file.json or use forward slashes instead. Invalid JSON syntax is another frequent culprit. Run your config file through a JSON validator before relaunching Claude.

If Claude connects but returns no Drive data, your OAuth token may have expired. Delete the cached token file (usually stored in the same directory as your credentials) and restart Claude to trigger a fresh authorization. For persistent errors, check the Claude Desktop logs found under Help > Open Logs in the menu bar.

What You Can Do Once It Is Running

A working MCP server connection transforms how you use Claude. You can ask it to summarize a long Google Doc, find all spreadsheets related to a project, pull data from a specific file, or compare two documents side by side. Claude handles those tasks through natural language. You do not need to copy and paste content manually.

The same pattern applies to other MCP servers. Adding the GitHub server lets Claude review pull requests and read code. The Slack server lets it summarize channel conversations. Each server you add expands what Claude can see and do on your behalf. That is the real power of the MCP architecture: it makes Claude a participant in your actual work environment, not just a standalone chatbot.

The Bottom Line

Setting up an MCP server with Claude is straightforward once you understand the pieces. The config file does the heavy lifting. The OAuth flow handles security. And the result is a version of Claude that actually knows your files, your projects, and your data.

Start with Google Drive to get comfortable with the process. Then explore the broader ecosystem of available MCP servers and keep expanding what Claude can do for you. The more context Claude has, the more useful it becomes.