A new command-line tool published on GitHub consolidates Workspace’s sprawling API into a single interface. This shows how seriously the company is taking the agentic AI moment.
The tool is described in its documentation as “one CLI for all Google Workspace, built for humans and AI agents.” GW. Provides integrated command-line access to Gmail, Google Drive, Calendar, Docs, Sheets, Slides, Chat, and most other workspace services.
But more obvious details are hidden in the instructions. The documentation includes a dedicated integration guide for OpenClaw, an open-source AI agent that went viral in late January and has since become something of a Rorschach test for where agent AI is headed.
Google’s decision to name-check OpenClaw in official documentation (even unofficial official documentation) was not made by the company by chance.
Why command-line tools are important for AI agents
EU technology 💜
The latest rumors from the EU tech industry, the story of the wise founder Boris, and some questionable AI art. Delivered free to your inbox every week. Sign up now!
in front G.W.S.An AI agent that searched Gmail inboxes, retrieved files from Drive, and updated calendar events had to navigate three separate APIs, each with its own authentication flow, rate limits, and response format. The process worked, but as PCWorld put it, it was “a huge pain.”
The new tool brings it all together in a single interface. All operations produce structured JSON output. This format is one that AI agents can reliably parse without any ambiguities that can hinder the functionality of graphical interfaces. Authentication is handled once by OAuth and then inherited by the agent that calls the tool.
This architecture has one particularly elegant feature. GW No static list of commands is provided. Instead, it reads Google’s own Discovery service at runtime and dynamically builds the entire command surface. When Google adds a new API endpoint, the tool automatically selects it.
No need to wrestle with versions that need to be updated or outdated documentation. For agents designed to operate over long periods of time, self-updating quality is less useful. That’s a meaningful guarantee of reliability.
The repository also includes over 100 pre-built “agent skills” covering common workspace workflows, such as uploading files to Drive with automatic metadata, adding data to spreadsheets, scheduling calendar events, forwarding Gmail attachments, and many similar operations.
These are separate, configurable building blocks that agent frameworks like OpenClaw are designed to chain together.
OpenClaw connection
The OpenClaw story is moving quickly. The project was published by Austrian software developer Peter Steinberger in November 2025 under the name Clawdbot, a name that prompted a trademark infringement complaint from Anthropic.
After a short stint as Moltbot, I settled on OpenClaw in late January 2026. Within weeks, users had created 1.5 million agents using the platform. The GitHub repository has accumulated nearly 200,000 stars. OpenClaw’s premise is simple enough to fit on a business card. It’s AI that actually does things.
On February 14, Sam Altman announced that Steinberger would be joining OpenAI to lead the next generation of personal agents. OpenClaw will move to an independent open source infrastructure supported by OpenAI. “Lobster is taking the world by storm,” Steinberger wrote in a farewell post. “My next mission is to build an agent that even mothers can use.”
The timing couldn’t be more coincidental that three weeks after Steinberger joined OpenAI, Google’s Workspace CLI appears in the middle of that story, with OpenClaw integration instructions in the documentation. It’s not confirmed whether it reflects a deliberate competitive reaction, a coincidental release, or simply whether Google’s developers shipped something that was already in the works.
What is clear is that major platform companies are now building the infrastructure to make their apps even more useful with the open source agent ecosystem that OpenAI just acquired Architects.
MCP and a wider field of view
Beyond OpenClaw GW Also acts as a model context protocol server. MCP is an open standard for how AI agents communicate with external tools, originally developed by Anthropic and now being adopted across the industry. running GWS MCP We expose the Workspace API as a structured tool that can be called natively by MCP-compatible clients, Claude Desktop, VS Code with AI extensions, or Google’s own Gemini CLI.
MCP support is important because it means this tool is more than just an OpenClaw utility. This is the infrastructure for an entire class of AI agents that are converging on MCP as the standard. Google is effectively making Workspace a first-class citizen of the emerging agent ecosystem, regardless of which model or framework is in play.
There is one important caveat. Google’s documentation clearly states: GW is “not an officially supported Google product.” Because this is published as a developer sample, there are no guarantees of stability, security, or ongoing maintenance at the production service level. For individual developers and experimenters, this is a manageable risk.
For companies considering deploying AI agents against live data in their workspaces, this is a meaningful limitation, especially given continued concerns about OpenClaw’s security model. Cisco researchers found that OpenClaw’s security model is vulnerable to data leakage and instant insertion by malicious third-party skills.
What Google is telling us
Google Cloud AI Director Addy Osmani is focusing her team on building an infrastructure for agent systems that can generate command-line input and manage structured output across complex workflows. Workspace CLI fits directly into that vision.
Wider patterns are easier to read. Microsoft has Copilot Tasks. OpenAI now has OpenClaw architects on board. Google has its own Gemini agent stack, with a CLI that makes today’s most widely used productivity suite readable by agents that speak JSON and MCP.
Competition over where enterprise AI agents can reside and what data they can access is accelerating, and the battleground is increasingly looking less like the applications themselves and more like the infrastructure underlying them.
For now, GW is a GitHub repository, but there are some caveats. But the fact that 14,000 stars were accumulated without most journalists realizing it suggests that the developers who make a living writing agents already understand what it means.
