Google Workspace Opens to AI Agents via the CLI

Opening Google Workspace to AI agents goes through the command line.

We’ve seen this approach lately with the arrival of several projects. For example, gogcli and workspace-cli. Both emerged at the end of 2025. They provide access to the core APIs of the suite, with a design tailored for AI agents.

Their future looks uncertain now that Google has its own tool: gws. As of now, it does not officially receive support. Yet the code has been pushed to the GitHub organization dedicated to Google Workspace.

Written in Rust like workspace-cli, gws also adapts the input format for AI agents, favoring raw JSON. It does the same with API responses; notably, to limit token consumption, it includes pagination and field masks.

Also read: Gemini in Google Workspace: announcements at a gallop, a rollout at a trot

The engineer who spearheaded it stresses the particular approach chosen for the API documentation. Rather than forcing a model to search online or embed API details into its system prompt, the API schemas are presented on request. The Google Workspace API discovery document—a dynamic catalog that describes the surface of the APIs and how to use them—acts as the single source of truth.

A hierarchy of skills… and a junction with OpenClaw

As with workspace-cli, gws enables the dry run (local validation of requests) and the control of API responses via Model Armor (a service hosted on Google Cloud).

Another shared feature was the availability of an MCP server, but that option ultimately disappeared from gws. As with multi-account support (agents could switch based on context).

The instructions for OpenClaw integrated into the README have remained. The skills system as well. There are roughly a hundred of them, spread across four levels of abstraction:

  • Services (the productivity core of Google Workspace, but also identity management, licenses, groups, and alerts, among other things)
  • Frequent operations (gws-mail-send, gws-drive-upload, etc.)
  • “Recipes” (combinations of operations: auditing shared Drive files, sending personalized emails based on a spreadsheet, etc.)
  • Personalities (product manager, sales, event coordinator, IT administrator…)
Dawn Liphardt

Dawn Liphardt

I'm Dawn Liphardt, the founder and lead writer of this publication. With a background in philosophy and a deep interest in the social impact of technology, I started this platform to explore how innovation shapes — and sometimes disrupts — the world we live in. My work focuses on critical, human-centered storytelling at the frontier of artificial intelligence and emerging tech.