Connect Claude AI to Google Sheets with an MCP Server

8/11/2025

Of course. Let's dive into how to build & connect your very own MCP Server. Turns out, this is a pretty powerful way to hook up AI like Claude directly to your Google Sheets, letting it read, write, & manage data for you.

Honestly, when I first heard "MCP Server," I thought it was some generic tech term. But it's actually something specific & pretty cool: the Model-Context-Protocol. Think of it as a standardized language that lets Large Language Models (LLMs) like Claude talk to other tools & services.

Instead of building a server from the ground up, the community has already created a fantastic one for Google Sheets. Our job is less about coding from scratch & more about getting the configuration right. It's like setting up a new piece of high-tech gear.

Here's the game plan:

The Google Cloud Setup: This is the most detailed part, where we create the credentials our server needs to talk to Google's APIs.
Running the MCP Server: We'll use a slick package manager to get the server running with a single command.
Connecting to Claude: We'll hook everything up to a client like Claude Desktop so you can start giving it commands.
How It All Works: We'll see how the Python code behind the server defines the tools Claude can use.

Let's get into it.

Part 1: The Foundation - Setting Up Your Google Cloud Credentials

This part is all about security & permissions. We need to create a special "Service Account" on Google Cloud Platform (GCP). This is like giving our server its own identity with specific permissions to access your files. It's WAY more secure than pasting your personal password into a script.

Here’s how you do it step-by-step:

Go to the Google Cloud Console: If you don't have a project, create a new one. Name it something like "My-MCP-Projects".
Enable the APIs: A service can't use an API if it's not turned on. We need two:
- In the console, search for "API & Services" -> "Library".
- Find & enable the Google Drive API.
- Find & enable the Google Sheets API.
Create the Service Account: This is the "robot" user that will act on your behalf.
- Navigate to "APIs & Services" -> "Credentials".
- Click "+ CREATE CREDENTIALS" & select "Service account".
- Give it a memorable name, like
  1mcp-sheets-controller
  . You'll see an email-like ID created for it.
Grant Permissions (The Important Part):
- When prompted to grant roles, give it the "Editor" role for now. This is a broad permission that lets it edit files. For a production system, you might want to get more granular, but for a personal project, Editor is the simplest path.
- Click "Continue" & "Done".
Generate a JSON Key: Now we need to give our server a secret key to log in as this service account.
- Back on the "Credentials" page, find the service account you just created. Click the three dots under "Actions" & select "Manage keys".
- Click "ADD KEY" -> "Create new key".
- Choose JSON as the type & click "CREATE".
- A
  1.json
  file will immediately download. TREAT THIS FILE LIKE A PASSWORD. Store it in a secure, memorable location on your computer. We'll need the path to this file in a minute.
Create a Google Drive Folder & Share It: The service account needs a place to work.
- Go to your regular Google Drive.
- Create a new folder. Call it something like "Claude-MCP-Work".
- Open the
  1.json
  file you downloaded. Find the
  1client_email
  field. It will look like
  1mcp-sheets-controller@your-project-name.iam.gserviceaccount.com
  . Copy this email address.
- Back in Google Drive, click the "Share" button on your new folder & paste in that service account email. Make sure to give it Editor permissions.
- Grab the Folder ID from the URL in your browser. It's the long string of characters after
  1.../folders/
  . Save this ID.

Phew, okay, that's the hardest part done. Now we have everything we need to actually run the server.

Part 2: Running the MCP Server with
`1uvx`

In the past, you'd have to clone a repository, install a bunch of Python dependencies, & run a script. It was a whole thing. Now, we can use a tool called

uv

from a company called Astral. It's a super-fast Python package installer.

uv

comes with a command,

uvx

, that can download & run a package in one go.

Install
1uv
: If you don't have it, open your terminal (or PowerShell on Windows) & run the installation command.
- macOS / Linux:
1curl -LsSf https://astral.sh/uv/install.sh | sh
- Windows:
1powershell -c "irm https://astral.sh/uv/install.ps1 | iex"
- Follow any instructions it gives you to add
  1uv
  to your system's PATH.
Set Environment Variables: Before we run the server, we need to tell it where to find our credentials & our working folder. These are the two pieces of info we saved from the last step.
- macOS / Linux:

Part 1: The Foundation - Setting Up Your Google Cloud Credentials

Part 2: Running the MCP Server with 1 uvx

Part 2: Running the MCP Server with
`1uvx`