A powerful Model Context Protocol (MCP) server enabling AI-native, secure, and extensible access to Google Workspace services
In the evolving landscape of AI-powered assistants, streamlined access to productivity tools is essential. The Google Workspace MCP Server, developed by Taylor Wilsdon, is a high-performance open-source server that bridges Google Workspace APIs (Calendar, Drive, Docs, Gmail) with Large Language Models (LLMs) via the Model Context Protocol (MCP).
This Python-based server is designed to empower AI assistants, autonomous agents, and chat-based tools to natively query, manipulate, and generate content using real-time, secure user data—without compromising on modularity, privacy, or developer control.
🔧 Key Features at a Glance
- OAuth 2.0 Authentication: Seamlessly manages user auth flows with secure token storage and refresh
- Google Calendar: Full CRUD operations for events with all-day/time-based support
- Google Drive: Search, read, and write files including native .docx/.xlsx extraction
- Gmail: Search inbox, retrieve messages, compose drafts, and send emails
- Google Docs: Create, read, and list documents—all accessible via LLMs
- Streamable HTTP + SSE: Dual transport support for compatibility with agents and chat UIs
- mcpo integration: Serve via OpenAPI and connect directly to tools like Open WebUI
- Extensible Design: Easily add support for more APIs or services
- Secure OAuth Callback Server: Built-in
/oauth2callbackhandling onlocalhost:8000 - Thread-safe session management for reliable multi-client concurrency
🔄 AI & LLM Integration Use Cases
This project is ideal for:
- LLM tools and copilots needing access to a user’s email, calendar, or documents
- Research agents generating reports from Google Docs and Drive files
- AI personal assistants managing schedule, writing emails, or summarizing documents
- Enterprise chatbots integrated into workflows with calendar or task visibility
It supports Open WebUI integration and is fully compatible with Claude Desktop, OpenRouter, or any MCP-compatible front-end.
⚙️ Quick Start for Sysadmins & Developers
Prerequisites
- Python 3.11+ and uv
- Google Cloud Project with OAuth 2.0 credentials
- Enabled APIs: Google Calendar, Drive, Gmail, Docs
Installation
git clone https://github.com/taylorwilsdon/google_workspace_mcp.git
cd google_workspace_mcp
uv venv
source .venv/bin/activate
uv pip install -e .
Place your client_secret.json in the root directory and never commit it.
Start the Server
export OAUTHLIB_INSECURE_TRANSPORT=1 # For local testing
uv run main.py
The server listens by default on http://localhost:8000.
🔌 mcpo + OpenAPI Integration
To expose the server as an OpenAPI tool:
{
"mcpServers": {
"google_workspace": {
"type": "streamablehttp",
"url": "http://localhost:8000/mcp"
}
}
}
Run mcpo:
mcpo --port 8001 --config config.json --api-key "secret"
Then connect via Swagger UI: http://localhost:8001/docs
Or as a tool in Open WebUI, using http://localhost:8001/google_workspace
📑 Available Tools (Sample)
Google Calendar
| Tool | Description |
|---|---|
list_calendars | Fetch all calendars |
get_events | Retrieve upcoming events |
create_event | Add new calendar events |
modify_event | Update existing events |
delete_event | Remove calendar entries |
Gmail
| Tool | Description |
|---|---|
search_gmail_messages | Use standard Gmail search syntax |
get_gmail_message_content | Retrieve full email content |
send_gmail_message | Send email |
draft_gmail_message | Compose a draft message |
Google Drive
| Tool | Description |
|---|---|
search_drive_files | Search by name, type, or content |
get_drive_file_content | Read specific file |
create_drive_file | Upload new document |
list_drive_items | Browse folders/files |
Google Docs
| Tool | Description |
|---|---|
search_docs | Find Docs by name |
get_doc_content | Read document text |
create_doc | Generate a new Doc |
All tools support session-based or email-based authentication. If no credentials exist, the assistant will prompt the user to complete OAuth via a secure redirect.
🔒 Security Notes
client_secret.jsonand user token files must be excluded from version control- Supports scoped OAuth requests—each tool only requests necessary access
- Consider reverse proxy (Nginx, Caddy) + API key for external production usage
🧩 Extendability
Want to add more Google Workspace tools?
- Define async tool in
gdrive/,gmail/, etc. - Use
@server.tool("your_tool_name") - Access credentials via
get_credentials() - Register the tool in
main.py
GitHub Repo: github.com/taylorwilsdon/google_workspace_mcp
License: MIT
Whether you’re building an AI assistant, LLM-based workflow, or autonomous agent, Google Workspace MCP Server offers a ready-to-deploy, developer-friendly bridge to integrate AI with real-world productivity tools.
