Google employees are testing an AI agent codenamed Remy inside an internal version of the Gemini app, according to reports published in early May. The system is described as a personal agent that works around the clock, watching a user’s calendar, email and deadlines and stepping in with suggestions or actions rather than waiting to be asked.
Google has not announced the product publicly, and the reports are based on people familiar with the internal trial. The company declined to comment in the coverage that first described the project.
What Remy is meant to do
The main distinction between Remy and the current Gemini assistant, according to the reports, is that Remy is proactive. If a meeting is coming up, emails are waiting for a reply or a deadline is approaching, the agent is designed to notice and act. It is also said to learn a user’s habits and preferences over time so that its help becomes more tailored.
That description fits a pattern across the industry. Over the past year, the leading AI companies have shifted their attention from chatbots that respond to prompts toward agents that can plan, take actions in software and run continuously on a user’s behalf.
Catching up in the agent race
Several outlets framed Remy as Google’s answer to OpenClaw, an open-source personal agent that gained a large following earlier in 2026 for running tasks across a user’s accounts and apps. OpenClaw’s creator joined OpenAI in February, and the project is widely cited as evidence that there is demand for always-on assistants that go beyond conversation.
Reports also say Google has started several other internal agent projects, including a coding tool reportedly known as Agent Smith, and that co-founder Sergey Brin, who has returned to hands-on work at the company, has pushed for agents to become central to Google’s AI strategy. Those details could not be independently confirmed.
Why it matters
Google has advantages that most agent developers do not: it already runs the calendar, email, documents and search services that a personal agent would need to work with, and Android gives it a path onto billions of phones. At the Android Show on May 12, the company publicly announced Gemini Intelligence, which lets Gemini carry out tasks across apps, signaling that at least part of this agent work is moving toward release.
The reports do not say when, or whether, Remy will reach the public. Internal codenames often change before launch, and features tested by employees are sometimes folded into other products. What the reporting does make clear is that Google sees autonomous agents, not chat, as the next competitive front.
