google-documents-mcp

MCP server for Google Docs - read and edit documents.

Use Cases

Document Q&A: "What does the contract say about termination clauses?" → reads the doc and pulls out the relevant section.

Meeting-to-doc workflow: After a call, reads the Gemini transcript plus follow-up Slack discussion, then updates your team charter with the decisions made.

Release announcement: "Draft a blog post about the new feature based on the GitHub PRs and Slack discussions in #proj-awesome" → synthesizes technical changes into user-facing content.

Daily planning with persistent context: Your AI assistant reads a "planning log" doc at the start of each day, asks about yesterday's progress, helps you identify today's priority, then updates the doc with your plan. Context carries over across sessions - it remembers what you said you'd do and can follow up.

(These are just examples - any workflow that needs document reading or editing can use this. Use in combination with google-drive-mcp for finding files, deleting, comments, and sharing permissions.)

Setup

$3

1. Go to Google Cloud Console
2. Create a new project (or use existing)
3. Enable the Google Docs API
4. Go to APIs & Services → OAuth consent screen, set up consent screen
5. Go to APIs & Services → Credentials → Create Credentials → OAuth client ID
6. Choose Web application
7. Add http://localhost:3000/callback to Authorized redirect URIs
8. Note your Client ID and Client Secret

$3

``bash
GOOGLE_CLIENT_ID='your-client-id' \
GOOGLE_CLIENT_SECRET='your-client-secret' \
MCP_TRANSPORT=http \
npm start
`

The server runs on http://localhost:3000 by default. Change with PORT=3001.

$3

`bash
claude mcp add --transport http google-documents-mcp http://localhost:3000/mcp
`

Architecture

This server acts as an OAuth proxy to Google:

`mermaid
graph LR
A[MCP client] <--> B[google-documents-mcp] <--> C[Google OAuth/API]
`

1. Server advertises itself as an OAuth authorization server via /.well-known/oauth-authorization-server
2. /register returns the Google OAuth client credentials
3. /authorize redirects to Google, encoding the client's callback URL in state
4. /callback receives the code from Google and forwards to the client's callback
5. /token proxies token requests to Google, injecting client credentials
6. /mcp handles MCP requests, using the bearer token to call Google Docs API

The server holds no tokens or state - it just proxies OAuth to Google.

Tools

| Tool | Description |
|------|-------------|
| document_get_raw | Get full raw JSON structure (all tabs, formatting, headers, footers, styles) |
| document_get_text | Get plain text content of all tabs |
| document_create | Create a new blank document |
| document_batch_update | Apply multiple edits atomically (insert, delete, format, etc.) |
| document_append | Append text to end of document |
| document_insert | Insert text at a specific index |
| document_replace | Find and replace text |

$3

The document_batch_update tool supports these operations:

- insertText - Insert text at a location
- deleteContentRange - Delete a range of content
- replaceAllText - Find and replace all occurrences
- insertInlineImage - Insert an image
- insertTable - Create a table
- insertTableRow / insertTableColumn - Add rows/columns to tables
- deleteTableRow / deleteTableColumn - Remove rows/columns from tables
- insertPageBreak - Add a page break
- createNamedRange / deleteNamedRange - Manage named ranges
- createParagraphBullets / deleteParagraphBullets - Manage bullet lists

Google Docs API Scopes

- documents - Full access to read and edit documents

Contributing

Pull requests are welcomed on GitHub! To get started:

1. Install Git and Node.js
2. Clone the repository
3. Install dependencies with npm install
4. Run npm run test to run tests
5. Build with npm run build

Releases

Versions follow the semantic versioning spec.

To release:

1. Use npm version to bump the version
2. Run git push --follow-tags` to push with tags
3. Wait for GitHub Actions to publish to the NPM registry.