Ultimate Google Docs & Drive MCP Server

Connect Claude Desktop (or other MCP clients) to your Google Docs and Google Drive!

🔥 Check out 15 powerful tasks you can accomplish with this enhanced server! 📁 NEW: Complete Google Drive file management capabilities!

This comprehensive server uses the Model Context Protocol (MCP) and the fastmcp library to provide tools for reading, writing, formatting, structuring Google Documents, and managing your entire Google Drive. It acts as a powerful bridge, allowing AI assistants like Claude to interact with your documents and files programmatically with advanced capabilities.

Features:

Document Access & Editing

• Read Documents: Read content with readGoogleDoc (plain text, JSON structure, or markdown)
• Append to Documents: Add text to documents with appendToGoogleDoc
• Insert Text: Place text at specific positions with insertText
• Delete Content: Remove content from a document with deleteRange
• Tab Support: Work with multi-tab documents using listDocumentTabs and optional tabId parameter in read/write operations

Formatting & Styling

• Text Formatting: Apply rich styling with applyTextStyle (bold, italic, colors, etc.)
• Paragraph Formatting: Control paragraph layout with applyParagraphStyle (alignment, spacing, etc.)
• Find & Format: Format by text content using formatMatchingText (legacy support)

Document Structure

• Tables: Create tables with insertTable
• Page Breaks: Insert page breaks with insertPageBreak
• Images: Insert images from URLs with insertImageFromUrl, or upload local images with insertLocalImage
• Experimental Features: Tools like fixListFormatting for automatic list detection

🆕 Comment Management

• List Comments: View all comments in a document with listComments (shows author, date, and quoted text)
• Get Comment Details: Get specific comment with replies using getComment
• Add Comments: Create new comments anchored to text with addComment
• Reply to Comments: Add replies to existing comments with replyToComment
• Resolve Comments: Mark comments as resolved with resolveComment
• Delete Comments: Remove comments from documents with deleteComment

🆕 Google Drive File Management

• Document Discovery: Find and list documents with listGoogleDocs, searchGoogleDocs, getRecentGoogleDocs
• Document Information: Get detailed metadata with getDocumentInfo
• Folder Management: Create folders (createFolder), list contents (listFolderContents), get info (getFolderInfo)
• File Operations: Move (moveFile), copy (copyFile), rename (renameFile), delete (deleteFile)
• Document Creation: Create new docs (createDocument) or from templates (createFromTemplate)

Integration

• Google Authentication: Secure OAuth 2.0 authentication with full Drive access
• MCP Compliant: Designed for use with Claude and other MCP clients
• VS Code Integration: Setup guide for VS Code MCP extension

---

Prerequisites

Before you start, make sure you have:

1. Node.js and npm: A recent version of Node.js (which includes npm) installed on your computer. You can download it from nodejs.org. (Version 18 or higher recommended).
2. Git: Required for cloning this repository. (Download Git).
3. A Google Account: The account that owns or has access to the Google Docs you want to interact with.
4. Command Line Familiarity: Basic comfort using a terminal or command prompt (like Terminal on macOS/Linux, or Command Prompt/PowerShell on Windows).
5. Claude Desktop (Optional): If your goal is to connect this server to Claude, you'll need the Claude Desktop application installed.

---

Setup Instructions

Follow these steps carefully to get your own instance of the server running.

Step 1: Google Cloud Project & Credentials (The Important Bit!)

This server needs permission to talk to Google APIs on your behalf. You'll create special "keys" (credentials) that only your server will use.

1. Go to Google Cloud Console: Open your web browser and go to the Google Cloud Console. You might need to log in with your Google Account.
2. Create or Select a Project:
   • If you don't have a project, click the project dropdown near the top and select "NEW PROJECT". Give it a name (e.g., "My MCP Docs Server") and click "CREATE".
   • If you have existing projects, you can select one or create a new one.
3. Enable APIs: You need to turn on the specific Google services this server uses.
   • In the search bar at the top, type "APIs & Services" and select "Library".
   • Search for "Google Docs API" and click on it. Then click the "ENABLE" button.
   • Search for "Google Drive API" and click on it. Then click the "ENABLE" button (this is often needed for finding files or permissions).
4. Configure OAuth Consent Screen: This screen tells users (usually just you) what your app wants permission for.
   • On the left menu, click "APIs & Services" -> "OAuth consent screen".
   • Choose User Type: Select "External" and click "CREATE".
   • Fill in App Information:
     • App name: Give it a name users will see (e.g., "Claude Docs MCP Access").
     • User support email: Select your email address.
     • Developer contact information: Enter your email address.
   • Click "SAVE AND CONTINUE".
   • Scopes: Click "ADD OR REMOVE SCOPES". Search for and add the following scopes:
     • https://www.googleapis.com/auth/documents (Allows reading/writing docs)
     • https://www.googleapis.com/auth/drive.file (Allows access to specific files opened/created by the app)
     • Click "UPDATE".
   • Click "SAVE AND CONTINUE".
   • Test Users: Click "ADD USERS". Enter the same Google email address you are logged in with. Click "ADD". This allows you to use the app while it's in "testing" mode.
   • Click "SAVE AND CONTINUE". Review the summary and click "BACK TO DASHBOARD".
5. Create Credentials (The Keys!):
   • On the left menu, click "APIs & Services" -> "Credentials".
   • Click "+ CREATE CREDENTIALS" at the top and choose "OAuth client ID".
   • Application type: Select "Desktop app" from the dropdown.
   • Name: Give it a name (e.g., "MCP Docs Desktop Client").
   • Click "CREATE".
6. ⬇️ DOWNLOAD THE CREDENTIALS FILE: A box will pop up showing your Client ID. Click the "DOWNLOAD JSON" button.
   • Save this file. It will likely be named something like client_secret_....json.
   • IMPORTANT: Rename the downloaded file to exactly credentials.json.
7. ⚠️ SECURITY WARNING: Treat this credentials.json file like a password! Do not share it publicly, and never commit it to GitHub. Anyone with this file could potentially pretend to be your application (though they'd still need user consent to access data).

Step 2: Get the Server Code

1. Clone the Repository: Open your terminal/command prompt and run:

   git clone https://github.com/a-bonus/google-docs-mcp.git mcp-googledocs-server

2. Navigate into Directory:

   cd mcp-googledocs-server

3. Place Credentials: Move or copy the credentials.json file you downloaded and renamed (from Step 1.6) directly into this mcp-googledocs-server folder.

Step 3: Install Dependencies

Your server needs some helper libraries specified in the package.json file.

1. In your terminal (make sure you are inside the mcp-googledocs-server directory), run:

   npm install

   This will download and install all the necessary packages into a node_modules folder.

Step 4: Build the Server Code

The server is written in TypeScript (.ts), but we need to compile it into JavaScript (.js) that Node.js can run directly.

1. In your terminal, run:

   npm run build

   This uses the TypeScript compiler (tsc) to create a dist folder containing the compiled JavaScript files.

Step 5: First Run & Google Authorization (One Time Only)

Now you need to run the server once manually to grant it permission to access your Google account data. This will create a token.json file that saves your permission grant.

1. In your terminal, run the compiled server using node:

   node ./dist/server.js

2. Watch the Terminal: The script will print:
   • Status messages (like "Attempting to authorize...").
   • An "Authorize this app by visiting this url:" message followed by a long https://accounts.google.com/... URL.
3. Authorize in Browser:
   • Copy the entire long URL from the terminal.
   • Paste the URL into your web browser and press Enter.
   • Log in with the same Google account you added as a Test User in Step 1.4.
   • Google will show a screen asking for permission for your app ("Claude Docs MCP Access" or similar) to access Google Docs/Drive. Review and click "Allow" or "Grant".
4. Get the Authorization Code:
   • After clicking Allow, your browser will likely try to redirect to http://localhost and show a "This site can't be reached" error. THIS IS NORMAL!
   • Look carefully at the URL in your browser's address bar. It will look like http://localhost/?code=4/0Axxxxxxxxxxxxxx&scope=...
   • Copy the long string of characters between code= and the &scope part. This is your single-use authorization code.
5. Paste Code into Terminal: Go back to your terminal where the script is waiting ("Enter the code from that page here:"). Paste the code you just copied.
6. Press Enter.
7. Success! The script should print:
   • "Authentication successful!"
   • "Token stored to .../token.json"
   • It will then finish starting and likely print "Awaiting MCP client connection via stdio..." or similar, and then exit (or you can press Ctrl+C to stop it).
8. ✅ Check: You should now see a new file named token.json in your mcp-googledocs-server folder.
9. ⚠️ SECURITY WARNING: This token.json file contains the key that allows the server to access your Google account without asking again. Protect it like a password. Do not commit it to GitHub. The included .gitignore file should prevent this automatically.

Step 6: Configure Claude Desktop (Optional)

If you want to use this server with Claude Desktop, you need to tell Claude how to run it.

1. Find Your Absolute Path: You need the full path to the server code.

   • In your terminal, make sure you are still inside the mcp-googledocs-server directory.
   • Run the pwd command (on macOS/Linux) or cd (on Windows, just displays the path).
   • Copy the full path (e.g., /Users/yourname/projects/mcp-googledocs-server or C:\Users\yourname\projects\mcp-googledocs-server).

2. Locate mcp_config.json: Find Claude's configuration file:

   • macOS: ~/Library/Application Support/Claude/mcp_config.json (You might need to use Finder's "Go" -> "Go to Folder..." menu and paste ~/Library/Application Support/Claude/)
   • Windows: %APPDATA%\Claude\mcp_config.json (Paste %APPDATA%\Claude into File Explorer's address bar)
   • Linux: ~/.config/Claude/mcp_config.json
   • If the Claude folder or mcp_config.json file doesn't exist, create them.

3. Edit mcp_config.json: Open the file in a text editor. Add or modify the mcpServers section like this, replacing /PATH/TO/YOUR/CLONED/REPO with the actual absolute path you copied in Step 6.1:

   {
     "mcpServers": {
       "google-docs-mcp": {
         "command": "node",
         "args": [
           "/PATH/TO/YOUR/CLONED/REPO/mcp-googledocs-server/dist/server.js"
         ],
         "env": {}
       }
       // Add commas here if you have other servers defined
     }
     // Other Claude settings might be here
   }

   • Make sure the path in "args" is correct and absolute!
   • If the file already existed, carefully merge this entry into the existing mcpServers object. Ensure the JSON is valid (check commas!).

4. Save mcp_config.json.

5. Restart Claude Desktop: Close Claude completely and reopen it.

---

Usage with Claude Desktop

Once configured, you should be able to use the tools in your chats with Claude:

• "Use the google-docs-mcp server to read the document with ID YOUR_GOOGLE_DOC_ID."
• "Can you get the content of Google Doc YOUR_GOOGLE_DOC_ID?"
• "Append 'This was added by Claude!' to document YOUR_GOOGLE_DOC_ID using the google-docs-mcp tool."

Working with Tabs

Google Docs now supports multi-tab documents. This MCP server provides full support for working with tabs:

Listing Tabs:

• "List all tabs in document YOUR_GOOGLE_DOC_ID using the listDocumentTabs tool."
• "Show me the tab structure with content summary for document YOUR_GOOGLE_DOC_ID."

Reading from Specific Tabs:

• "Read the content from tab TAB_ID in document YOUR_GOOGLE_DOC_ID using the readGoogleDoc tool."
• "Get the markdown content from tab TAB_ID in document YOUR_GOOGLE_DOC_ID."

Writing to Specific Tabs:

• "Append 'New content' to tab TAB_ID in document YOUR_GOOGLE_DOC_ID."
• "Insert text at index 100 in tab TAB_ID of document YOUR_GOOGLE_DOC_ID."
• "Delete content from range 50-100 in tab TAB_ID of document YOUR_GOOGLE_DOC_ID."

Note: The following tools support the optional tabId parameter:

• readGoogleDoc - Read from a specific tab
• appendToGoogleDoc - Append to a specific tab
• insertText - Insert text into a specific tab
• deleteRange - Delete content from a specific tab

When tabId is not specified, operations target the first tab (or the legacy document body for older documents without tabs).

Advanced Usage Examples:

• Text Styling: "Use applyTextStyle to make the text 'Important Section' bold and red (#FF0000) in document YOUR_GOOGLE_DOC_ID."
• Paragraph Styling: "Use applyParagraphStyle to center-align the paragraph containing 'Title Here' in document YOUR_GOOGLE_DOC_ID."
• Table Creation: "Insert a 3x4 table at index 500 in document YOUR_GOOGLE_DOC_ID using the insertTable tool."
• Image Insertion: "Use insertImageFromUrl to insert an image from 'https://example.com/image.png' at index 100 in document YOUR_GOOGLE_DOC_ID."
• Local Image Upload: "Use insertLocalImage to upload '/path/to/image.jpg' and insert it at index 200 in document YOUR_GOOGLE_DOC_ID."
• Legacy Formatting: "Use formatMatchingText to find the second instance of 'Project Alpha' and make it blue (#0000FF) in doc YOUR_GOOGLE_DOC_ID."

Remember to replace YOUR_GOOGLE_DOC_ID with the actual ID from a Google Doc's URL (the long string between /d/ and /edit).

Claude will automatically launch your server in the background when needed using the command you provided. You do not need to run node ./dist/server.js manually anymore.

---

Image Insertion

This server provides two ways to insert images into Google Documents:

1. Insert from Public URL (insertImageFromUrl)

Inserts an image directly from a publicly accessible URL. The image URL must be accessible without authentication.

Parameters:

• documentId: The Google Document ID
• imageUrl: Publicly accessible URL (http:// or https://)
• index: Position in the document (1-based indexing)
• width (optional): Image width in points
• height (optional): Image height in points

Example:

"Insert an image from https://example.com/logo.png at index 100 in document YOUR_DOC_ID"

2. Upload Local Image (insertLocalImage)

Uploads a local image file to Google Drive and inserts it into the document. This is a two-step process that:

1. Uploads the image to Google Drive (by default to the same folder as the document)
2. Makes the image publicly readable
3. Inserts the image into the document using its Drive URL

Parameters:

• documentId: The Google Document ID
• localImagePath: Absolute path to the local image file
• index: Position in the document (1-based indexing)
• width (optional): Image width in points
• height (optional): Image height in points
• uploadToSameFolder (optional, default: true): If true, uploads to the document's folder; if false, uploads to Drive root

Supported formats: .jpg, .jpeg, .png, .gif, .bmp, .webp, .svg

Example:

"Upload and insert the image at /Users/myname/Pictures/chart.png at index 200 in document YOUR_DOC_ID with width 400 and height 300"

Note: The uploaded image will be made publicly readable so it can be displayed in the document. The image file will remain in your Google Drive and can be managed separately.

---

Security & Token Storage

• .gitignore: This repository includes a .gitignore file which should prevent you from accidentally committing your sensitive credentials.json and token.json files. Do not remove these lines from .gitignore.
• Token Storage: This server stores the Google authorization token (token.json) directly in the project folder for simplicity during setup. In production or more security-sensitive environments, consider storing this token more securely, such as using system keychains, encrypted files, or dedicated secret management services.

---

Testing

The multi-tab support features have been thoroughly tested and verified:

✅ Tested Features:

• listDocumentTabs - Lists all tabs with IDs, titles, positions, and content summaries
• readGoogleDoc with tabId - Reads specific tabs; backward compatible without tabId
• appendToGoogleDoc with tabId - Appends to specific tabs without affecting others
• insertText with tabId - Inserts text at specific positions in specific tabs
• deleteRange with tabId - Deletes content from specific tabs in isolation
• Multi-tab operations - Sequential operations on different tabs work independently
• Error handling - Invalid tab IDs return clear, helpful error messages
• Backward compatibility - Operations without tabId default to first tab (legacy documents supported)

All tab-related features have been validated with real Google Docs containing multiple tabs, confirming:

• Tab isolation (operations on one tab don't affect others)
• Proper tab ID validation and error messages
• Correct content retrieval and manipulation per tab
• Full backward compatibility with single-tab and legacy documents

Troubleshooting

• Claude shows "Failed" or "Could not attach":
  • Double-check the absolute path in mcp_config.json.
  • Ensure you ran npm run build successfully and the dist folder exists.
  • Try running the command from mcp_config.json manually in your terminal: node /PATH/TO/YOUR/CLONED/REPO/mcp-googledocs-server/dist/server.js. Look for any errors printed.
  • Check the Claude Desktop logs (see the official MCP debugging guide).
  • Make sure all console.log status messages in the server code were changed to console.error.
• Google Authorization Errors:
  • Ensure you enabled the correct APIs (Docs, Drive).
  • Make sure you added your email as a Test User on the OAuth Consent Screen.
  • Verify the credentials.json file is correctly placed in the project root.
• Tab-related Errors:
  • If you get "Tab with ID not found", use listDocumentTabs to see all available tab IDs
  • Ensure you're using the correct tab ID format (typically a short alphanumeric string)
  • Single-tab documents don't require tabId - operations work on the document body automatically

---

License

This project is licensed under the MIT License - see the LICENSE file for details. (Note: You should add a LICENSE file containing the MIT License text to your repository).