Google preps Gemini agent for macOS to control computers and organize files, challenging Claude Cowork
Google is developing a Gemini agent for macOS that will control computers, organize files, and integrate with Google Workspace apps. Code analysis reveals features including file conversion to Google Sheets, folder organization, batch file renaming, and meeting follow-up automation.
Google Preps Gemini Agent for macOS to Control Computers and Organize Files
Google is developing a Gemini agent for macOS with computer control capabilities to rival Anthropic's Claude Cowork, according to code analysis by 9to5Google's APK Insight team.
The agent will use macOS Screen Access and Accessibility features to control mouse and keyboard inputs, similar to Claude Cowork's computer use functionality released in recent months.
Four Core Capabilities Revealed
Code strings reveal four example prompts that define the agent's initial scope:
- Convert files to spreadsheets: Scan local folders for invoices or reports, extract data, and structure it into Google Sheets
- Organize folders: Find unorganized files in Desktop or Downloads, group by type or context, and archive
- Standardize files: Read file metadata and batch-rename files into organized subfolders
- Meeting follow-up: Extract Meet transcripts or Docs notes from recent meetings and draft follow-up emails with highlights and action items
The first three capabilities focus on local file management and Google Workspace integration. The fourth shifts focus to Google's productivity suite including Meet, Docs, and Gmail.
Current Gemini macOS App Limited
The existing Gemini macOS app offers two features: a native chat interface and an alt-space shortcut that allows sharing the current window for visual context. The upcoming agent capabilities would represent a significant expansion.
Broader Than Android Implementation
The macOS agent appears positioned for broader capabilities than Gemini's Android implementation. Currently, only select Android devices like the Galaxy S26 series can automate simple in-app tasks such as food ordering.
Google previously demonstrated computer control capabilities in the Gemini 2.5 Computer Use preview released last year, but has not yet shipped production features.
What This Means
Google is playing catch-up to Anthropic's Claude Cowork in the computer control agent space, focusing initially on file management and Google Workspace integration rather than general-purpose automation. The emphasis on Google Workspace suggests the company plans to leverage its productivity suite as a competitive advantage, potentially making the agent more valuable for organizations already committed to Google's ecosystem. No release timeline has been announced.
Related Articles
Google Expands Gemini-Powered Ask Maps Globally With Personal Intelligence, Real-Time Transit, Agentic Ordering
Google Maps' Gemini-powered Ask Maps chat is rolling out globally to English speakers in Australia, Brazil, Canada, Indonesia, Japan, Mexico, and over 150 other countries and territories. The update adds Gmail-based Personal Intelligence, real-time transit data, conversation memory, and agentic capabilities for ordering food and booking hotels.
Anthropic Adds Cross-Session Messaging to Claude Code v2.1.224
Claude Code v2.1.224 introduces cross-session messaging, letting separate Claude Code instances on macOS and Linux send each other summaries to coordinate work. The feature does not support approving permissions or executing commands remotely.
Amazon, Cursor, Microsoft, OpenAI, and Vercel Launch Agent Plugins, a Shared Packaging Standard for AI Agent Extensions
Amazon, Cursor, Microsoft, OpenAI, and Vercel have released Agent Plugins, an open standard defining a single package format for AI agent extensions. Version 1.0.0 covers Agent Skills and MCP servers, but leaves marketplaces, permissions, and runtime out of scope.
AWS Adds Temporal Policies to Bedrock AgentCore to Stop AI Agents From Exploiting Multi-Step Actions
Amazon Bedrock AgentCore now supports temporal policies—stateful authorization rules that evaluate an AI agent's entire session history, not just individual tool calls. The feature runs at the AgentCore Gateway, outside agent code, so it cannot be bypassed by prompt manipulation or agent bugs.
Comments
Loading...