Google Workspace MCP Server
A hardened, safety-first Google Calendar MCP server enabling AI agents to read and manage calendars with strict data isolation and runtime security limits.
README
Google Workspace MCP Server
A hardened, safety-first Google Calendar Model Context Protocol (MCP) server designed for enterprise reliability, strict data isolation, and deep agentic workflow integrations. This server empowers AI agents (like Claude Desktop, Cursor, or Antigravity) to read and manage calendars while enforcing strict user-defined security limits at runtime.
โจ Core Features & Safety Constraints
To protect user accounts and prevent unauthorized modifications, the server implements multiple runtime safeguards:
- ๐ Domain Boundary Whitelist: Blocks agents from sending calendar invites to external or unauthorized email domains (via
GWS_ALLOWED_DOMAINS). - ๐ก๏ธ PII & Description Scrubber: Redacts email addresses, phone numbers, and meeting details dynamically based on privacy levels (
GWS_PII_MODE). - ๐ฅ Double-Booking Prevention: Checks calendar availability automatically before booking an event unless the client explicitly requests an override (
ignore_availability=True). - โ ๏ธ Two-Turn Delete Flow: Forces a dry-run confirmation step for event deletions, returning a detailed preview card before executing changes.
- ๐ Profile-Based Token Storage: Isolates authentication tokens based on
GWS_PROFILEto allow multi-tenant/account setups (e.g., separating personal and corporate accounts). - โฑ๏ธ API Rate Limiter: Enforces a global minimum interval of 100ms between calls to avoid flooding Google Cloud API quotas.
- ๐ Structured Audit Logging: Appends timestamped, owner-only readable (
chmod 600) JSON log lines for all modifying actions.
๐ Quick Start & Installation
1. Run the Interactive Installer (Recommended)
We provide an interactive installer script that handles virtual environments, fetches dependencies, guides you through placing your Google credentials, runs the auth flow, and auto-injects configuration settings into Claude Desktop:
# Clone the repository and kick off the configuration engine
curl -fsSL https://raw.githubusercontent.com/wrik0/gws-workspace-agent/master/setup.sh | bash
Alternatively, if you already cloned the repository locally:
./setup.sh
You can also configure a custom profile (e.g., work) during setup:
./setup.sh --profile work
2. Manual Installation
If you prefer to configure the environment manually:
# Clone the repository
git clone https://github.com/wrik0/gws-workspace-agent.git
cd gws-workspace-agent
# Set up environment and install package in editable mode
uv venv
uv pip install -e ".[dev]"
3. Google OAuth Setup
Before running the server, place your Google Cloud Client Secrets file (credentials.json) in the standard application data folder:
- Linux:
~/.config/google-workspace-mcp/credentials.json - macOS:
~/Library/Application Support/google-workspace-mcp/credentials.json - Windows:
%APPDATA%\gws\google-workspace-mcp\credentials.json
Next, run the CLI authorization utility:
# Performs browser login and caches the OAuth token
# On Unix (using the symlinked binary):
gws-auth
# Or directly via virtual environment:
.venv/bin/gws-auth
โ๏ธ Environment Configuration
You can configure the server behaviour using environment variables or equivalent command-line options when serving:
| Environment Variable | CLI Option | Default | Description |
|---|---|---|---|
GWS_PROFILE |
--profile |
default |
Workspace profile name. Resolves to token_{GWS_PROFILE}.json. |
GWS_MODE |
--mode / --readonly |
full |
Mode of operation: full (read/write tools) or readonly (read tools only). |
GWS_PII_MODE |
--pii-mode |
redact |
Privacy filter: none (no masking), redact (phone/emails masked), metadata_only (description/title removed). |
GWS_ALLOWED_DOMAINS |
--allowed-domains |
None | Comma-separated domain whitelist for invitees (e.g. company.com,partner.org). |
GWS_TIMEZONE |
None | System Local | Local timezone database name (e.g. America/New_York or Asia/Kolkata). |
GWS_AUDIT_LOG |
None | APP_DIR/audit.log |
Destination path for the append-only JSON audit logs. |
๐ Command Line Interface (CLI)
The package exposes three entry points:
gws-serve
Starts the stdio FastMCP server.
gws-serve --mode readonly --pii-mode redact --allowed-domains company.com
gws-auth
Starts the OAuth token refresh browser loop.
# Request only read-only scopes from Google
gws-auth --mode readonly
# Request write scopes (requires verification if app is in sandbox)
gws-auth --mode full
gws-purge
Purges cached tokens and credentials files.
# Deletes cached token file but leaves credentials.json intact
gws-purge --token-only
# Deletes the entire application config directory
gws-purge --all
๐ Client Integration
Claude Desktop
Add this to your claude_desktop_config.json:
{
"mcpServers": {
"google-workspace": {
"command": "~/.local/bin/gws-serve",
"args": ["--readonly", "--pii-mode", "redact"],
"env": {
"GWS_ALLOWED_DOMAINS": "mycompany.com"
}
}
}
}
Cursor IDE
Navigate to Cursor Settings > Features > MCP > + Add New MCP Server:
- Name:
google-workspace - Type:
command - Command:
~/.local/bin/gws-serve --readonly --pii-mode redact
๐ฆ MCP Tools Reference
Read Tools (Available in readonly and full modes)
list_calendars(): Lists all calendars in the account with their name, ID, and access roles (e.g. owner, reader).view_schedule(time_min, time_max, calendar_id="primary"): Displays events in a specific window. Outputs summary, times, descriptions, and color emojis (derived from Google Color IDs).find_slots(time_min, time_max, duration_minutes, calendar_id="primary"): Queries free/busy state and aggregates open slots available for scheduling.
Write Tools (Available only in full mode)
create_event(summary, start_iso, end_iso, ...): Creates a calendar event. Validates invitee domains, checks double-booking conflicts, and records the action inaudit.log.move_event(event_id, new_start_iso, new_end_iso): Shift events in time via a singlePATCHAPI request.modify_event(event_id, ...): Dynamically updates text details or appends new whitelisted attendees.delete_event(event_id, confirm=False): Two-turn deletion flow. Run withconfirm=Falseto inspect the event details card, and run withconfirm=Trueto execute the API call.
๐งช Developer & Testing Suite
We use pytest for unit testing. The test suite uses mocks to fully simulate Google Cloud services and config directories:
# Run pytest locally
.venv/bin/pytest -v
Style & Format Enforcement
To maintain code quality, run Ruff checkers:
# Format codebase
.venv/bin/ruff format
# Run linter checks
.venv/bin/ruff check --fix
๐ License
This project is licensed under the terms of the MIT license.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.