mcp-google

README

Google MCP Server

An MCP (Model Context Protocol) server for reading Google Slides presentations and Google Docs documents.

Features

Read-only access to Google Slides and Google Docs via MCP tools.

Google Slides Tools

• get_presentation - Get presentation metadata (title, slide count, locale, revision ID)
• list_slides - List all slides with titles and element counts
• get_slide - Get slide content and element details by index or object ID
• get_slide_notes - Get speaker notes by slide index or object ID
• get_presentation_content - Get all slides' text and images in one call (ideal for AI)

Google Docs Tools

• get_document - Get document metadata (title, word count, element counts)
• get_document_content - Get structured content (headings, paragraphs, images, tables)
• get_document_text - Get all text as a single plain text string
• get_document_paragraphs - Get text organized by paragraphs

Architecture

This server is built on omniskill, making its Google Slides and Docs skills composable building blocks that can be reused in multi-service MCP servers.

Composable Skills

The skills in this repository can be imported and combined with other skills:

import (
    "github.com/grokify/mcp-google/skills/slides"
    "github.com/grokify/mcp-google/skills/docs"
    runtime "github.com/plexusone/omniskill/mcp/server"
)

// Create runtime
rt := runtime.New(&mcp.Implementation{
    Name:    "work-mcp-server",
    Version: "v1.0.0",
}, nil)

// Add Google skills
slidesSkill := slides.New(googleHTTPClient)
slidesSkill.Init(ctx)
rt.RegisterSkill(slidesSkill)

docsSkill := docs.New(googleHTTPClient)
docsSkill.Init(ctx)
rt.RegisterSkill(docsSkill)

// Add other skills (Slack, Jira, GitHub, etc.)
rt.RegisterSkill(slackSkill)
rt.RegisterSkill(jiraSkill)

// Run server
rt.ServeStdio(ctx)

This enables building unified MCP servers that combine multiple services while keeping each service's implementation modular and maintainable.

Requirements

• Go 1.24+
• Google Cloud service account with Slides and Docs API access

Installation

go install github.com/grokify/mcp-google/cmd/mcp-google@latest

Or build from source:

git clone https://github.com/grokify/mcp-google.git
cd mcp-google
go build ./cmd/mcp-google

Setup

1. Create a Google Cloud Service Account

1. Go to the Google Cloud Console
2. Create a new project or select an existing one
3. Enable the Google Slides API and Google Docs API
4. Create a service account with no special roles
5. Download the JSON credentials file

2. Share Documents with the Service Account

Share any presentations or documents you want to access with the service account's email address (found in the credentials JSON as client_email).

Usage

Option 1: Google Service Account Credentials

Use a standard Google Cloud service account JSON file:

mcp-google --credentials /path/to/service-account.json

Or using an environment variable:

export GOOGLE_CREDENTIALS_FILE=/path/to/service-account.json
mcp-google

Option 2: goauth CredentialsSet

Use a goauth CredentialsSet file, which can store multiple credentials:

mcp-google --goauth-credentials-file /path/to/credentials.json --goauth-credentials-account myaccount

Or using environment variables:

export GOAUTH_CREDENTIALS_FILE=/path/to/credentials.json
export GOAUTH_CREDENTIALS_ACCOUNT=myaccount
mcp-google

The CredentialsSet entry should be of type gcpsa with appropriate scopes:

{
  "credentials": {
    "myaccount": {
      "type": "gcpsa",
      "gcpsa": {
        "gcpCredentials": {
          "type": "service_account",
          "project_id": "...",
          "private_key_id": "...",
          "private_key": "...",
          "client_email": "...",
          "client_id": "..."
        },
        "scopes": [
          "https://www.googleapis.com/auth/presentations.readonly",
          "https://www.googleapis.com/auth/documents.readonly",
          "https://www.googleapis.com/auth/drive.readonly"
        ]
      }
    }
  }
}

Option 3: Vault-Backed Credentials

Use omnitoken with omnivault-desktop for secure credential storage in password managers.

Supported vault providers:

Provider	URI Pattern	Requirements
1Password	op://vault	OP_SERVICE_ACCOUNT_TOKEN env var
Bitwarden	bw://org-id	BW_ACCESS_TOKEN and BW_ORGANIZATION_ID env vars
File	file:///path	None
Env	env://PREFIX_	None

1Password Example

export OP_SERVICE_ACCOUNT_TOKEN="ops_..."
mcp-google --vault op://MyVault --credentials-name google

Bitwarden Example

export BW_ACCESS_TOKEN="..."
export BW_ORGANIZATION_ID="..."
mcp-google --vault bw://org-id --credentials-name google

File Vault Example

mcp-google --vault file:///path/to/secrets --credentials-name google

Claude Desktop Configuration

Add to your Claude Desktop configuration (claude_desktop_config.json):

With Google Service Account

{
  "mcpServers": {
    "google": {
      "command": "/path/to/mcp-google",
      "env": {
        "GOOGLE_CREDENTIALS_FILE": "/path/to/service-account.json"
      }
    }
  }
}

With goauth CredentialsSet

{
  "mcpServers": {
    "google": {
      "command": "/path/to/mcp-google",
      "env": {
        "GOAUTH_CREDENTIALS_FILE": "/path/to/credentials.json",
        "GOAUTH_CREDENTIALS_ACCOUNT": "myaccount"
      }
    }
  }
}

With 1Password

{
  "mcpServers": {
    "google": {
      "command": "/path/to/mcp-google",
      "env": {
        "OP_SERVICE_ACCOUNT_TOKEN": "ops_...",
        "OMNITOKEN_VAULT_URI": "op://MyVault",
        "OMNITOKEN_CREDENTIALS_NAME": "google"
      }
    }
  }
}

See docs/configuration/claude-desktop.md for more options including Bitwarden.

Google Slides Tools

get_presentation

Get metadata about a presentation.

Input:

• presentation_id (required) - The ID of the Google Slides presentation

Output:

• title - Presentation title
• slide_count - Number of slides
• locale - Presentation locale
• revision_id - Current revision ID

list_slides

List all slides in a presentation.

Input:

• presentation_id (required) - The ID of the Google Slides presentation

Output:

• slides - Array of slide information:
  • object_id - Slide's unique identifier
  • index - Zero-based slide index
  • title - Slide title (if present)
  • element_count - Number of elements on the slide

get_slide

Get the content of a specific slide.

Input:

• presentation_id (required) - The ID of the Google Slides presentation
• slide_index (optional) - Zero-based slide index
• slide_object_id (optional) - Slide's object ID

One of slide_index or slide_object_id must be provided.

Output:

• text_content - Array of text strings from the slide
• element_summary - Array of element details:
  • object_id - Element's unique identifier
  • element_type - Type of element (shape, image, table, etc.)
  • description - Element description or text preview

get_slide_notes

Get the speaker notes for a specific slide.

Input:

• presentation_id (required) - The ID of the Google Slides presentation
• slide_index (optional) - Zero-based slide index
• slide_object_id (optional) - Slide's object ID

One of slide_index or slide_object_id must be provided.

Output:

• notes - Speaker notes text

get_presentation_content

Get all slide content in a single call - ideal for AI analysis of the entire presentation.

Input:

• presentation_id (required) - The ID of the Google Slides presentation
• include_notes (optional) - Include speaker notes for each slide (default: false)

Output:

• title - Presentation title
• slides - Array of slide content:
  • index - Zero-based slide index
  • object_id - Slide's unique identifier
  • title - Slide title (if present)
  • text_content - Array of text strings from the slide
  • images - Array of images:
    • object_id - Image element ID
    • content_url - Direct URL to image (valid ~30 minutes)
    • source_url - Original source URL (if available)
    • alt_text - Image description
  • notes - Speaker notes (if include_notes is true)

Google Docs Tools

get_document

Get metadata about a document.

Input:

• document_id (required) - The ID or URL of the Google Doc

Output:

• title - Document title
• document_id - Document ID
• revision_id - Current revision ID
• word_count - Approximate word count
• char_count - Character count
• image_count - Number of images
• table_count - Number of tables
• header_count - Number of headers
• footer_count - Number of footers

get_document_content

Get the full structured content of a document.

Input:

• document_id (required) - The ID or URL of the Google Doc
• include_images (optional) - Include image information (default: false)
• include_tables (optional) - Include table content (default: false)
• include_headers (optional) - Include document headers (default: false)
• include_footers (optional) - Include document footers (default: false)

Output:

• title - Document title
• sections - Array of content sections:
  • type - Section type ("heading", "paragraph")
  • level - Heading level (1-6, for headings only)
  • text - Section text content
  • style_id - Style identifier (e.g., "HEADING_1", "NORMAL_TEXT")
• images - Array of images (if requested):
  • object_id - Image element ID
  • content_uri - Direct URL to image
  • source_uri - Original source URL
  • title - Image title
  • description - Image description
• tables - Array of tables (if requested):
  • rows - Number of rows
  • columns - Number of columns
  • cells - 2D array of cell text content
• headers - Array of header text (if requested)
• footers - Array of footer text (if requested)

get_document_text

Get all text from a document as a single plain text string.

Input:

• document_id (required) - The ID or URL of the Google Doc

Output:

• title - Document title
• text - Full document text

get_document_paragraphs

Get text organized by paragraphs.

Input:

• document_id (required) - The ID or URL of the Google Doc

Output:

• title - Document title
• paragraphs - Array of paragraph text strings

Finding Document IDs

Presentations

The presentation ID is in the URL when viewing a presentation:

https://docs.google.com/presentation/d/PRESENTATION_ID_HERE/edit

Documents

The document ID is in the URL when viewing a document:

https://docs.google.com/document/d/DOCUMENT_ID_HERE/edit

Note: Google Docs tools accept either the document ID or the full URL, including URLs with query strings and anchors:

https://docs.google.com/document/d/DOCUMENT_ID_HERE/edit?tab=t.0#heading=h.xyz

License

MIT