Google Workspace CLI: The Missing Tool for Vibe Coders Building AI Agents

Google just dropped a tool that vibe coders have been waiting for: gws, a CLI that gives you (and your AI agents) full access to Google Workspace — Gmail, Docs, Sheets, Drive, Calendar, and more — from the terminal or directly through Cursor's MCP integration.

npx @googleworkspace/cli

That's it. One command, and you have access to over 100 skills across the entire Google Workspace API.

What is gws?

The Google Workspace CLI (@googleworkspace/cli) is an open-source tool built by Justin Poehnelt from Google's Workspace team. It's written in Rust for speed and compiled to a single binary — no Python environments, no Java runtimes, no dependency hell.

What makes it different from other Google API wrappers:

100+ skills out of the box — Gmail, Docs, Sheets, Drive, Calendar, Admin, and more

Auto-adapts to new APIs — as Google ships new Workspace APIs, gws picks them up automatically

Built-in MCP server — works natively with Cursor, Claude Code, and any MCP-compatible tool

AI-native design — skills are structured so LLMs can discover and use them without custom prompting

The project hit 6,100 GitHub stars within days of launch, was publicly praised by Addy Osmani (Chrome team lead) and Guillermo Rauch (Vercel CEO), and is already being used by developers building AI agents on top of Workspace.

Why This Matters for Vibe Coders

If you're building apps with Cursor and your users live in Google Workspace (most businesses do), you've probably hit this wall: connecting your app to Gmail, Docs, or Sheets means wading through Google's OAuth setup, API client libraries, and scope management. It's the kind of work that breaks the vibe.

gws eliminates that friction. Instead of writing Google API boilerplate, you describe what you want:

"Read my Gmail and summarize unread emails"

"Create a Google Doc from this app output"

"Pull data from this spreadsheet and use it in my app"

"Add a calendar event for every deployment"

Your AI agent handles the API calls, authentication, and data formatting.

Setting Up gws as an MCP Server in Cursor

Step 1: Install and authenticate

Run the CLI once to authenticate with your Google account:

npx @googleworkspace/cli auth login

This opens a browser window for OAuth. Grant the permissions you need (Gmail read, Docs write, etc.) and you're set.

Step 2: Add the MCP config to Cursor

Add this to your .cursor/mcp.json:

{
  "mcpServers": {
    "google-workspace": {
      "command": "npx",
      "args": ["@googleworkspace/cli", "mcp"]
    }
  }
}

Step 3: Start prompting

Open Cursor's agent mode and try these:

Read and summarize email: "Read my last 10 unread Gmail messages and give me a summary of what needs my attention today."

Create a doc from code output: "Take the API documentation we just generated and create a Google Doc called 'API Reference v2' in my project folder on Drive."

Pull spreadsheet data into your app: "Read the pricing data from my 'Product Catalog' spreadsheet and generate a TypeScript const with the data."

Automate meeting notes: "Check my Google Calendar for today's meetings and create a blank Google Doc for each one titled 'Notes: [meeting name] — March 5, 2026'."

Real Example: Building an AI Agent That Uses Workspace

Here's a practical scenario. You're building a SaaS app with Cursor, and you want an onboarding flow that:

Creates a shared Google Drive folder for each new customer

Copies your template docs into that folder

Sends a welcome email from your business Gmail

Adds a kickoff meeting to your Google Calendar

Without gws, this is a multi-day integration project — OAuth flows, service accounts, API client setup, scope management, and error handling for each Google API.

With gws as an MCP server in Cursor, you describe the flow in natural language and the agent builds it. The MCP skills handle authentication, API discovery, and execution. You focus on the product logic.

The Technical Details

Built in Rust. The CLI compiles to a native binary, so it starts instantly and uses minimal memory. No JVM warm-up, no interpreter overhead. When your AI agent calls a gws skill, the response comes back fast enough that it doesn't bottleneck the conversation.

Auto-adapting skills. Google ships API updates frequently. Instead of manually adding support for new endpoints, gws reads from Google's Discovery Documents and generates skills dynamically. When Google adds a new Workspace API, gws supports it without a release.

100+ skills for AI tools. The skill system is designed for LLM consumption. Each skill has structured input/output schemas that Claude, Gemini, and GPT can parse without ambiguity. This isn't a CLI that happens to work with AI — it's an AI-native tool that happens to have a CLI interface.

How It Compares

The key difference: gws is the only option that gives your AI agent direct, structured access to Google Workspace. Zapier can trigger actions, but it can't read your email and reason about it. API clients can do everything, but they require you to write the integration code. gws does both — read and write — and your AI agent drives it.

Who's Behind It

Justin Poehnelt is a developer on Google's Workspace team who has been building developer tools for Google APIs for years. This isn't a side project that might disappear — it's maintained by someone inside Google who understands the API surface deeply.

The early reception tells the story: 6,100 stars on GitHub within days, endorsements from Addy Osmani (who called it "the future of CLI tools") and Guillermo Rauch (who highlighted its MCP integration). The developer community is treating this as infrastructure, not hype.

Our Take

We use Google Workspace for everything at gptsters.com — docs, email, spreadsheets for data tracking. Before gws, connecting our AI workflows to Workspace meant writing custom API integrations. Now we add one MCP server to Cursor and it just works.

If you're building AI agents that need to interact with real business data, most of that data lives in Google Workspace. gws makes that data accessible to your agents without writing integration code.

Bottom line: Install it, authenticate once, add the MCP config to Cursor, and start building AI agents that can read, write, and manage Google Workspace. It takes 5 minutes to set up and saves days of API integration work.

npx @googleworkspace/cli

Further Reading

Set up your vibe coding environment

Compare AI coding tools

How MCP servers work

gws on GitHub