GWS Connector

Multi-account Google Workspace MCP server — connect multiple Gmail, Google Calendar, and Google Drive accounts with smart routing.

Works with Claude Code, Gemini CLI, GitHub Copilot, Cursor, OpenAI Codex, and any MCP-compatible client.

Why

Most AI coding assistants support a single Google account. If you use multiple Google accounts (personal + work, multiple clients, different orgs), you need to switch between them manually. This MCP server lets you connect them all at once and route requests by label, email, or domain.

Related MCP server: google-workspace-mcp

Features

• Multi-account — connect unlimited Gmail and Google Workspace accounts

• Smart routing — target accounts by label (work), email, or domain

• Per-account OAuth — different orgs can use their own GCP credentials

• Secure storage — client secrets and tokens stored in OS keychain (file fallback on Linux without GNOME Keyring)

• 30 tools — Mail (9), Calendar (4), Drive (3), Sheets (4), Docs (4), account management (6)

• Account management — add, remove, set default, list accounts

• Cross-platform — standard MCP server works with any compatible client

How it compares

There are several good Google Workspace MCP servers. GWS Connector is the one to pick when multiple accounts and operational simplicity matter:

If you only ever use a single Google account and want the widest possible tool surface (Slides, Forms, Chat, Tasks, …), a single-account server like taylorwilsdon/google_workspace_mcp may fit better. GWS Connector focuses on doing multi-account Gmail / Calendar / Drive / Sheets / Docs cleanly and securely.

Upgrading from v0.2.x

v0.3.0 adds native Google Sheets and Google Docs tools behind two new OAuth scopes (spreadsheets, documents). Existing users must re-authorize each connected account so new tokens are minted with these scopes:

/gws:reauth

Before approving the browser consent screen, review what the new scopes grant — full read and write access to every spreadsheet and document in that account's Google Drive, including files shared with the account. See the scope rationale table below for details.

You must also add the two new scopes (and enable the Sheets and Docs APIs) in your GCP project's OAuth consent screen configuration before re-auth, or the consent screen will reject the request.

Quick Start (Claude Code)

1. Install the plugin — run these two commands inside Claude Code:

/plugin marketplace add orieg/gws-connector
/plugin install gws@gws-connector

2. Set up Google Cloud credentials — the interactive wizard walks you through everything:

/gws:configure

This creates a GCP project, enables APIs, and connects your first account (~5 minutes). See Google Cloud Setup if you prefer manual steps.

3. Connect additional accounts:

/gws:add-account

Each account can use different OAuth credentials from different GCP projects.

Gemini CLI

gemini extensions install https://github.com/orieg/gws-connector

The binary is downloaded automatically on first use. Then connect accounts inside Gemini:

gws.accounts.add(label: "personal", clientId: "your-client-id", clientSecret: "your-secret")

Other clients

Download a prebuilt binary or build from source:

git clone https://github.com/orieg/gws-connector && cd gws-connector && make build

Then configure your client:

Connect accounts via MCP tool call:

gws.accounts.add(label: "personal", clientId: "your-client-id", clientSecret: "your-secret")

Environment variables (all optional): GWS_GOOGLE_CLIENT_ID, GWS_GOOGLE_CLIENT_SECRET, GWS_STATE_DIR

The --use-dot-names flag uses gws.mail.search naming; without it, tools use gws_mail_search.

git clone https://github.com/orieg/gws-connector
cd gws-connector
make build
claude --plugin-dir ./

Use /reload-plugins inside the session after making changes. Run claude --debug --plugin-dir ./ to troubleshoot plugin loading.

Each release attaches a one-click gws-mcp.mcpb bundle. Download it and open it with Claude Desktop (Settings → Extensions → install from file), or drag it in. The bundle contains the binaries for macOS and Linux and picks the right one for your machine automatically. You still complete the Google Cloud setup and connect accounts on first use.

The server is also published to the official MCP Registry as io.github.orieg/gws-connector, so MCP clients that browse the registry can find and install it directly.

A multi-arch image is published to GHCR on each release:

docker run -i --rm ghcr.io/orieg/gws-connector:latest

The server speaks MCP over stdio. Interactive OAuth (accounts.add / reauth) opens a browser and stores secrets in the OS keychain, so it needs host access — day-to-day use is best via the native binary, the Claude Code plugin, or the Gemini extension. The image is well suited to headless stdio integrations and to registry/introspection checks. Persist the account registry across runs by mounting a volume and pointing GWS_STATE_DIR at it:

docker run -i --rm -v gws-state:/state -e GWS_STATE_DIR=/state \
  ghcr.io/orieg/gws-connector:latest

Usage

All gws.* tools accept an optional account parameter:

# Uses default account
gws.mail.search(q: "is:unread")

# Target by label
gws.cal.list_events(account: "work")

# Target by email
gws.drive.search(account: "alice@company.com", q: "quarterly report")

Available tools

Skills

Interactive workflows available in both Claude Code and Gemini CLI:

Recipes

Once accounts are connected, just ask your assistant in plain language — it picks the tools and the account. Examples:

• Morning triage across accounts — "Summarize my unread email from the last 24 hours across all accounts, grouped by account, and flag anything that needs a reply today."

• Draft a reply in a thread — "Find the thread with Acme about the Q3 invoice on my work account and draft a reply confirming the new date. Don't send it."

• Turn an email into a calendar event — "Read the latest message from the events team and create a calendar event on my personal calendar with the date and location from it."

• Cross-account digest — "What meetings do I have tomorrow across my work and personal calendars? List them in one timeline."

• Find and summarize a doc — "Search my client-acme Drive for the latest 'statement of work' and give me the key deliverables and dates."

• Log to a spreadsheet — "Append a row to the 'Expenses' sheet in my personal Drive: today's date, 'AWS', 42.50."

• Keep inbox tidy — "Label all unread messages from newsletters@ as 'Newsletters' and mark them read on my personal account."

Tips:

• Target an account explicitly with its label ("on my work account"), by email, or by domain — otherwise the default account is used.

• Write operations (drafts, events, sheet/doc edits) are previewed for your confirmation before anything is sent or changed.

Google Cloud Setup

One-time setup (~5 minutes):

1. Go to Google Cloud Console and create a new project (e.g., "GWS Connector")

2. Enable APIs — click each link and hit "Enable":

   • Gmail API

   • Calendar API

   • Drive API

   • Sheets API

   • Docs API

3. Configure the OAuth consent screen:

   • Choose "External" (or "Internal" for Google Workspace orgs)

   • Fill in the app name (e.g., "Claude GWS") and your email for support contact

   • Click "Save"

4. Add scopes — go to Data Access:

   • Click "Add or Remove Scopes"

   • Add these 7 scopes (paste into the "Manually add scopes" box):

     • https://www.googleapis.com/auth/gmail.modify

     • https://www.googleapis.com/auth/calendar

     • https://www.googleapis.com/auth/drive

     • https://www.googleapis.com/auth/spreadsheets

     • https://www.googleapis.com/auth/documents

     • https://www.googleapis.com/auth/userinfo.email

     • https://www.googleapis.com/auth/userinfo.profile

   • Click "Update", then "Save"

   Why each scope is requested:

   Scope	Purpose	Tools
   gmail.modify	Read, draft, modify messages and labels	gws.mail.*
   calendar	Read and create/update events	gws.cal.*
   drive	Search and read files and metadata across Drive	gws.drive.*
   spreadsheets	Read and write Google Sheets cell data and metadata	gws.sheets.*
   documents	Read and write Google Docs content	gws.docs.*
   userinfo.email	Identify the authorizing account (email match on reauth)	account management
   userinfo.profile	Store a display name alongside the email	account management

5. Add test users — go to Audience:

   • Add each Google email address you plan to connect

   • ⚠️ This is required — without this you'll get "Access blocked: has not completed the Google verification process" (error 403) during OAuth

6. Create OAuth credentials — go to Clients:

   • Click "+ Create Client" → "OAuth client ID"

   • Application type: Desktop app

   • Click "Create"

   • Download the JSON file (click the download icon) — this contains your Client ID and Client Secret

Multiple organizations

If you connect accounts from different Google Workspace orgs, each org needs its own GCP project. Create OAuth credentials in each project and provide them when connecting:

gws.accounts.add(label: "work", clientId: "work-client-id", clientSecret: "work-secret")
gws.accounts.add(label: "personal", clientId: "personal-client-id", clientSecret: "personal-secret")

Client secrets are stored in the OS keychain. Client IDs are stored in the account registry.

Architecture

gws-connector/
├── cmd/gws-mcp/                 # MCP server entrypoint
├── internal/
│   ├── accounts/                # Account registry & router
│   ├── auth/                    # OAuth flow, token store, client factory
│   ├── server/                  # MCP tool registration & dispatch
│   └── services/                # Gmail, Calendar, Drive API wrappers
│
├── .claude-plugin/              # Claude Code plugin manifest + marketplace
├── .mcp.json                    # Claude Code MCP config
├── gemini-extension.json        # Gemini CLI extension manifest
├── CONTEXT.md                   # Shared behavioral context (both agents)
├── skills/                      # Slash commands (Claude Code + Gemini CLI)
├── hooks/                       # Claude Code session hooks
├── agents/                      # Claude Code workspace agent
│
├── .vscode/mcp.json             # GitHub Copilot MCP config
├── .cursor/mcp.json             # Cursor MCP config
└── codex.json                   # OpenAI Codex CLI config

• Token storage: OS keychain (macOS Keychain, GNOME Keyring, Windows Credential Manager) with automatic file fallback

• Client secrets: OS keychain per account (not stored in config files)

• Account registry: JSON file at ~/.claude/channels/gws/accounts.json (contains client IDs and metadata, no secrets)

• Credential resolution: per-account credentials (keychain) → global env var fallback

• Protocol: MCP (Model Context Protocol) over stdio — compatible with any MCP client

Development

make build          # Build binary
make test           # Run tests with race detector
make test-verbose   # Run tests with verbose output
make lint           # Run go vet
make release        # Cross-compile for all platforms
make clean          # Remove build artifacts

License

MIT — see LICENSE.