LABSCO SUMMARY
The service-account path is what makes it usable unattended.
Most Workspace servers assume a person is present to complete a browser sign-in. This one takes a service-account key on the command line instead, which is what a scheduled job or a server-side agent needs. The flip side of one binary covering seven services is scope: the OAuth consent you grant covers everything you enabled, so grant only the APIs you actually intend it to touch.
What it is
A single Go binary that exposes seven Google Workspace services through one MCP server. Rather than running a separate server per service, you install one command and it covers mail, files, calendar, spreadsheets, documents, contacts and tasks.
What you get
- Gmail — search and list threads, read full conversations, create drafts, send, and move to trash
- Drive — search, read text content, create files and folders, update content, move, share and trash
- Calendar — list upcoming events, create meetings with attendees, delete events
- Sheets — create spreadsheets, read ranges, append rows, update specific cells
- Docs — create new documents and read full document text
- People — list contacts and create new connections
- Tasks — list task lists and tasks, create, update and delete, with optional status or due-date filtering
Requirements
Go 1.24 or later, then go install github.com/matheusbuniotto/go-google-mcp/cmd/go-google-mcp@latest. Two authentication paths. For personal use, create a Desktop App OAuth client in Google Cloud, save it as client_secrets.json, and run go-google-mcp auth login --secrets <YOUR_PATH_TO_CLIENT_SECRETS> once — the token is stored in ~/.go-google-mcp/. For server or automated use, download a service-account key and pass -creds <YOUR_CREDS>. Enable the APIs you intend to use (Drive, Gmail and so on) in the same Google Cloud project. Clients launch it as go-google-mcp with no arguments.
Setup effort
Follow the published steps — go install github.com/matheusbuniotto/go-google-mcp/cmd/go-google-mcp@latest
USE CASES
Finding a thread and drafting the reply
Search Gmail, read the whole conversation, and leave a draft for you to check before it goes.
Putting a meeting in the calendar from a chat
Create the event with attendees directly instead of switching to Calendar.
Logging something to a spreadsheet
Append a row or update a specific cell in an existing sheet.
Running Workspace tasks from a server with no browser
Authenticate with a service-account key rather than an interactive login.
BEFORE YOU INSTALL
What to know first
Requires
- Go 1.24 or later
- go-google-mcp on PATH
INSTALLATION
Connect your client
Server setup
go install github.com/matheusbuniotto/go-google-mcp/cmd/go-google-mcp@latestIn the Google Cloud Console (https://console.cloud.google.com/), enable the APIs you need (Drive, Gmail and so on) and create a Desktop App OAuth client.
Download its client secrets file and save it as client_secrets.json.
Sign in once; the token is saved to ~/.go-google-mcp/:
go-google-mcp auth login --secrets <YOUR_PATH_TO_CLIENT_SECRETS>For server or automated use, skip the sign-in and run it with a service-account key instead: go-google-mcp -creds <YOUR_CREDS>
MCP Clients
Claude Desktop, Cursor
{
"mcpServers": {
"google-workspace": {
"command": "go-google-mcp",
"args": []
}
}
}Gemini CLI
gemini mcp add google-workspace $(which go-google-mcp)