Justin Poehnelt spent nearly seven years at Google. He was on the Workspace Developer Relations team, a role where building open-source layers and abstractions over Google APIs was, by his own description, what the team regularly did. Two months ago, he was fired. The reason, he says, was a CLI tool he built for Google Workspace that went viral before anyone had time to think about what to do with it.

Poehnelt disclosed the firing this week in a post on X, and the story has since picked up significant steam. The tool in question, gws, is a command-line interface that gives users and AI agents unified access to Google Drive, Gmail, Calendar, and every other Workspace API. What makes it technically interesting is that it doesn’t ship with a hardcoded list of commands. Instead, it reads Google’s own Discovery Service at runtime and dynamically builds its entire command surface from there. When Google adds or modifies an API endpoint, gws picks it up automatically — no manual updates required. The tool also shipped with 40+ pre-written agent skill files, structured JSON output, and zero boilerplate, making it genuinely useful for both developers and AI agents in the way few tools in this space are.

It was built, as Poehnelt described in a blog post, agents-first from day one: “Not ‘built a CLI, then noticed agents were using it.’ From Day One, the design assumptions were shaped by the fact that AI agents would be the primary consumers of every command, every flag, and every byte of output.”

When the tool launched in early March, it spread fast. It hit number one on Hacker News, gained thousands of GitHub stars, and drew many thousands of actual users within a couple of days. Google leaders and directors, Poehnelt says, were reaching out to ask what they could learn from it.

Then came the other side. Legal grilled him about why Google’s logo and brand colors appeared on a Google Workspace GitHub repository — a question whose irony is somewhat hard to miss given that the project lived under the official googleworkspace GitHub organization.

Poehnelt was fired shortly after. His read on why: “I think the cause was that Workspace and certain leaders (and projects) were afraid of being disrupted. But the fear wasn’t specific to my CLI, it was a broader fear in what agents meant for Workspace.”

The timing was particularly pointed. Two days before he was let go, Google announced at Google Cloud Next 2026 that an official Workspace CLI was in development. Google Cloud CEO Thomas Kurian used the event to declare that “the era of the pilot is over. The era of the agent is here.” And yet the engineer who had already built exactly that — an agent-ready, dynamically generated Workspace CLI — was being walked out the door.

The tweet that Poehnelt refers to as “the tweet that got me fired” came from Addy Osmani, then a Director at Google Cloud AI, who shared the tool on March 5 with the description: “Introducing the Google Workspace CLI: built for humans and agents.” Osmani, a 14-year Google veteran who led developer experience across Chrome and later Google Cloud AI, has since left the company as well.

There’s a legitimate debate about the branding question. Some commenters on Hacker News have noted that releasing a project with Google’s logo without going through brand approval was a policy violation regardless of where the project was hosted or how technically impressive it was. That’s a fair point, and Poehnelt hasn’t disputed that the process wasn’t followed. What he’s pushing back on is the interpretation that the branding issue was the real reason, rather than the cover for one.

The broader context matters here. Google has been openly working through what AI agents mean for its Workspace business. Gemini Spark — the company’s push to make Gemini an always-on agent with access to Gmail, Calendar, Drive, and more — is its consumer-facing answer to that question. But agents that can interact directly with Workspace APIs through a CLI represent a different kind of threat: one that gives developers and enterprise teams a way to build Workspace integrations without going through Google’s sanctioned surfaces, and one that could theoretically let other AI systems do what Gemini is supposed to do natively.

Sundar Pichai has acknowledged publicly that Google is “a bit behind” in agentic coding with tool use — one of the most commercially significant AI battlegrounds right now. An engineer inside Google building the exact kind of agent-ready tooling the company was behind on, and doing it faster and with more community uptake than internal teams had managed, puts an uncomfortable mirror in front of leadership.

Poehnelt says he wanted the story out there. “I want this out there because it is easier for me to explain my story and it is an experience I want to fully own. It’s also part of my healing.” He was clear that his time at Google gave him things he’s grateful for — strong teammates, a manager who backed him through the months of uncertainty — but he wasn’t willing to let the story remain unexplained.

The gws CLI, for what it’s worth, is still available on GitHub under the googleworkspace organization, under the Apache-2.0 license, with a note at the top that it is “not an officially supported Google product.” Whether Google eventually ships its own version that does what Poehnelt’s already does — or whether the official CLI announced at Cloud Next quietly becomes the same thing — remains to be seen.