Hardened Google Workspace MCP
A security-hardened integration that enables Claude to interact with Google Workspace tools like Gmail, Drive, and Docs while preventing data exfiltration by disabling features like email sending and external file sharing. It allows users to safely manage emails, documents, and calendars through natural language without risking unauthorized external communication.
README
Hardened Google Workspace MCP
A security-hardened Google Workspace integration for Claude Code.
This is a fork of taylorwilsdon/google_workspace_mcp with dangerous operations removed to prevent data exfiltration via prompt injection attacks. See SECURITY.md for details.
What This Does
Enables Claude Code to interact with your Google Workspace:
- Gmail: Read emails, create drafts (cannot send)
- Google Drive: Read/create files (cannot share externally)
- Google Docs: Read and edit documents
- Google Sheets: Read and write spreadsheets
- Google Calendar: View and create events
- Google Forms: Read and create forms
- Google Slides: Read and edit presentations
Why "Hardened"?
LLMs are vulnerable to prompt injection attacks—malicious instructions hidden in content the model processes. An attacker could embed instructions in an email or document that trick the AI into exfiltrating sensitive data.
This fork removes all tools that could send data outside your account:
- No email sending - Claude can draft emails, but you must manually send them from Gmail
- No file sharing - Claude cannot share files with external users
- No filter creation - Claude cannot create auto-forwarding rules
- Secure credential storage - OAuth tokens stored in macOS Keychain, not plaintext files
See SECURITY.md for the complete security model.
Prerequisites
- Claude Code installed on your machine
- A Google Workspace or personal Google account
- Python 3.11+ installed
Quick Start
Step 1: Create OAuth Credentials
Follow OAUTH_SETUP.md to create Google Cloud OAuth credentials.
Step 2: Install Dependencies
cd ~/hardened-google-workspace-mcp # or wherever you placed it
uv sync
Note: If you don't have
uvinstalled, run:curl -LsSf https://astral.sh/uv/install.sh | sh
Step 3: Configure Claude Code
Add the MCP server using claude mcp add:
claude mcp add hardened-workspace \
--scope user \
-e GOOGLE_OAUTH_CLIENT_ID="YOUR_CLIENT_ID" \
-e GOOGLE_OAUTH_CLIENT_SECRET="YOUR_CLIENT_SECRET" \
-- uv run --directory ~/hardened-google-workspace-mcp python -m main --single-user
Replace YOUR_CLIENT_ID and YOUR_CLIENT_SECRET with your OAuth credentials.
Or manually add to ~/.claude/mcp_config.json:
{
"mcpServers": {
"hardened-workspace": {
"command": "uv",
"args": ["run", "--directory", "/Users/YOUR_USERNAME/hardened-google-workspace-mcp", "python", "-m", "main", "--single-user"],
"env": {
"GOOGLE_OAUTH_CLIENT_ID": "YOUR_CLIENT_ID",
"GOOGLE_OAUTH_CLIENT_SECRET": "YOUR_CLIENT_SECRET"
}
}
}
}
Step 4: Authorize with Google
- Start (or restart) Claude Code
- The first time you use a Google Workspace tool, a browser window will open
- Sign in with your Google account
- Click "Allow" to grant permissions
For detailed instructions, see SETUP.md.
Example Prompts
Once set up, try these prompts in Claude Code:
List my recent emails from the past week
Read the document "Q4 Planning" from my Google Drive
Create a draft email to john@example.com about the meeting tomorrow
Show me what's on my calendar for next Monday
Update cell A1 in my "Budget 2025" spreadsheet to "Updated"
Troubleshooting
"OAuth credentials not found"
Make sure you've set the GOOGLE_OAUTH_CLIENT_ID and GOOGLE_OAUTH_CLIENT_SECRET environment variables in your MCP config.
"Permission denied" errors
- Delete the credentials folder:
rm -rf ~/.credentials/workspace-mcp/ - Restart Claude Code
- Re-authorize with your Google account
"Tool not found" errors
Make sure the MCP server is running. Check Claude Code's MCP status panel.
Browser doesn't open for authorization
If the browser doesn't open automatically, check the Claude Code output for a URL to copy/paste manually.
Security Notes
- Never disable permission prompts - Always review what Claude is asking to do
- Drafts require manual sending - Claude can create email drafts, but you must open Gmail to send them
- No external sharing - Claude cannot share files outside your organization
- Report issues - If Claude behaves unexpectedly, file an issue
See SECURITY.md for comprehensive security documentation.
Support
For issues with this project, please file an issue on GitHub.
Based on google_workspace_mcp by Taylor Wilsdon, licensed under MIT.
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.
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.
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.
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.