GWS CLI: The Google Workspace Tool That Isn't Officially Google (But Feels Like It Should Be)

Google just shipped a command-line interface for the entire Google Workspace suite. You can now control Gmail, Drive, Calendar, Docs, Sheets, Slides, Chat, and Admin from your terminal—or hand all of that access to an AI agent. The project lives at googleworkspace/cli on GitHub. It already has 10,000 stars in its first week.

There's a catch: Google officially calls it "not officially supported." And yet it's built by Google, open-sourced by Google, and is already one of the most interesting tools for AI-powered Workspace automation I've seen in years.

TL;DR: GWS CLI (gws) is a unified command-line tool for all Google Workspace APIs. It's dynamically built from Google's own API schema, ships with 100+ AI agent skills, and includes native MCP server support. I tested it and the quality of generated documents and presentations is noticeably better than anything else I've tried. It's not official, but it's promising. Watch this one closely.

What Is GWS CLI?

The project's full name is the Google Workspace CLI, and the command is gws. It's written in Rust, distributed via npm, and does something clever that most CLI tools don't: instead of hardcoding a list of commands, it reads Google's own Discovery Service at runtime and builds its entire command surface dynamically.

That means when Google adds a new API endpoint to any Workspace product, gws picks it up automatically on the next invocation. No manual updates. No waiting for a new release. The tool stays in sync with the platform by design.

The command surface covers essentially all of Google Workspace:

Gmail — read, compose, search, label, archive
Drive — list, upload, download, share, organize
Calendar — create events, query availability, manage invites
Docs & Slides — create, read, and modify documents and presentations
Sheets — read and write spreadsheet data
Chat — send messages, manage spaces
Admin — manage users, groups, and org settings

All of this from a single CLI. All of it accessible to an AI agent.

"Not Official" but Almost Official

Let's address the elephant in the room. The repository is at github.com/googleworkspace/cli. That's the official Google Workspace GitHub organization. The contributors are Google engineers. The license is Apache 2.0. The tool is built from Google's own Discovery Service.

And yet the docs explicitly state: "not officially supported."

In practice, this means Google isn't committing to a stable API contract, a support SLA, or long-term maintenance guarantees. Breaking changes can and will happen—as they already have: at one point, the team deleted 1,151 lines of MCP server code as a breaking change, then rebuilt the architecture from scratch with a smarter approach.

It's built by Google. It's open-sourced by Google. It's just not supported by Google. That's a meaningful distinction—but it's not a reason to ignore it.

The project has the DNA of an internal tool that got open-sourced. That's actually a good sign: internal tools are built to solve real problems, not to hit a product roadmap. The fact that it exists in this form suggests the team is dogfooding it.

The AI Agent Story

What makes gws genuinely interesting isn't just the CLI itself—it's the AI agent architecture baked into it.

The repository ships with over 100 agent skills (structured SKILL.md files) covering individual APIs and higher-level workflows. There are 50+ curated recipes for common tasks across Gmail, Drive, Calendar, Docs, and Sheets. These skills are designed to be dropped directly into agent environments like Claude Code, OpenClaw, or any MCP-compatible framework.

Speaking of MCP: gws runs as a native Model Context Protocol server, which means you can connect any MCP-compatible AI directly to all of Google Workspace through natural language tool calls. The team went through an interesting architectural evolution here. They initially built a full MCP implementation, realized it exposed 200–400 tools (well beyond what most clients can handle), then refactored to a compact mode that exposes one tool per service plus a gws_discover meta-tool. The result: 26 tools instead of 400, with 2–8x better practical performance in MCP clients.

That kind of thoughtful engineering decision tells you a lot about the team behind this.

My Experience: The Output Quality Is Different

I've tested a lot of tools that claim to generate Google Docs and Slides. Most of them fall into one of two failure modes: the output is structurally correct but looks generic and lifeless, or the formatting is a mess that requires significant cleanup before it's usable.

gws produces something different. The documents and presentations I generated felt closer to what a competent human would produce—not because the AI was smarter, but because the tool has direct, structured access to the Google Slides and Docs APIs. There's no conversion layer, no copy-paste from a third-party format, no "export as." The AI writes directly into the native format using the native API.

The difference shows. Layouts are consistent. Fonts and spacing follow Workspace defaults. Slide hierarchies make sense. It's the kind of output where your first reaction is "wait, did a human do this?"—rather than "okay, let me spend 20 minutes fixing this."

I'm still in early testing. I haven't stress-tested it across complex documents or long workflows. But the initial results are genuinely encouraging in a way that other tools in this space haven't been.

What It Doesn't Do (Yet)

A few honest caveats from my testing and from reading the project carefully:

No stability guarantees. The "not officially supported" label means you should expect breaking changes, especially as the MCP architecture continues to evolve.
Authentication setup is non-trivial. You'll need to configure OAuth credentials for the APIs you want to use. It's well-documented, but it's not a one-click install.
The agent skill library is still growing. 100+ skills sounds like a lot, but for complex, multi-step Workspace workflows, you'll likely find gaps that require custom work.
No GUI.** This is a terminal tool. If your team isn't comfortable with CLIs or agent configuration, the adoption curve will be steep.

None of these are dealbreakers. They're just the realistic state of a tool that was released last week and is actively being developed.

Why This Matters Beyond the CLI

The real significance of gws isn't that you can now type commands to control Gmail. It's what it signals about where Google is heading with Workspace and AI.

For the past few years, the AI-in-productivity-tools story has been about magic buttons inside the app: Gemini in Docs, Copilot in Word, AI-generated email replies. These are useful, but they're fundamentally bounded by the UI. You can't compose them. You can't automate them at scale. You can't hand them to an agent and say "go run this workflow while I focus on something else."

gws represents a different model: Workspace as a programmable platform, accessible to both humans and AI agents through a unified interface. That's a fundamentally different capability, and it opens workflows that weren't previously possible:

An agent that monitors your Gmail, extracts action items, creates Docs summaries, and schedules Calendar blocks—automatically
CI/CD pipelines that write deployment reports directly to Google Docs and notify teams via Chat
Automated slide deck generation from data in Sheets, triggered by a single command or agent call
Bulk administrative workflows that would take hours in the UI, running in seconds from a script

These aren't hypothetical. These are use cases that become straightforward once you have a reliable, unified CLI with AI agent support.

Should You Use It Today?

If you're a developer or technically comfortable user who lives in Google Workspace, yes—start experimenting now. The tool is already functional, the quality is surprisingly good, and getting familiar with it early means you'll be ahead of the curve when it stabilizes.

If you're evaluating it for production use in a team environment, wait a few months. Let the breaking changes settle. Watch how the MCP architecture evolves. The project is moving fast, and that's both its strength and its current limitation.

This is one of those tools you want to keep on your radar—not because it's finished, but because of what it's clearly becoming.

Google has spent years building the most widely used productivity suite in the world. gws is the first real sign that they're opening that suite to programmatic control at the level AI agents need. The "not officially supported" label will either get dropped as the project matures, or it will get forked by the community into something that lasts regardless.

Either way: watch this space.

GWS CLI is available at github.com/googleworkspace/cli. Install via npm install -g @googleworkspace/cli.