Google Unveils Gemini Cloud Agent: Autonomous AI Coworker with Claude Support and Cross-Platform Integration

On October 9, 2026, Google officially introduced Gemini Cloud Agent (AI Coworker), a next-generation autonomous collaborative agent system that unifies conversational chat, knowledge documentation, software development, and media generation into a single endpoint.

Image source: @Fei2411 / Google

Moving beyond isolated chatbots and passive prompt assistants, this release establishes a persistent 'Coworker Agent' architecture equipped with dedicated email identities and persistent storage, while expanding model selection beyond Google's proprietary Gemini line to include Anthropic's Claude models.

Unified API Integration with Cross-Platform and Headless Background Execution

The newly unveiled Gemini Cloud Agent consolidates previously fragmented workflows across chatting, knowledge management, code development, and multimedia generation into one agent architecture and unified API.

• Cross-Platform Sync: Sessions maintain seamless state continuity across web browsers, mobile devices (iOS and Android), and desktop environments (Windows and Mac).
• Multi-Channel Workspaces: Direct connectivity extends across terminal CLI tools, Google Workspace (Gmail, Docs, Sheets), Microsoft 365, and Slack to share collaborative context.
• Headless Background and Scheduling: Agents can operate as persistent headless services embedded within third-party applications, supporting recurring scheduled tasks (cron) and event-driven autonomous execution.

By shifting agent runtimes to a cloud-managed persistent infrastructure, the system overcomes traditional local session limitations and enables headless execution for long-running, recurring enterprise workflows.

Autonomous Coworker Identities and Dynamic Sub-Agent Orchestration

A key architectural pillar of Gemini Cloud Agent is its persistent 'Coworker Agent' structure.

Each agent can be provisioned with an independent identity, including a dedicated email address under the @agents.com domain, a synced calendar, and isolated persistent storage, allowing it to communicate and coordinate schedules alongside human team members. For large-scale or multi-stage initiatives, the agent dynamically spawns and manages subordinate sub-agents to delegate and execute parallel tasks under unified multi-agent orchestration.

In tandem with multi-agent coordination, the platform incorporates a persistent memory layer that retains interaction history and user feedback across sessions to refine task execution over time.

Multi-Model Claude Support and Automated Spending Caps

On the infrastructure and execution tier, Google has broadened model access beyond its internal catalog.

• Multi-Model Support: In addition to standard Gemini foundation models, users can select Anthropic's Claude family depending on task requirements, with plans to incorporate additional open-source and commercial proprietary models sequentially.
• Spending Cap Safeguards: To mitigate cost overruns during autonomous execution loops, administrators can enforce budget thresholds (spending caps) that immediately pause agent operations when financial limits are met.

This architecture offers development and operations teams greater flexibility in choosing specialized inference models while maintaining strict budgetary governance over autonomous cloud workflows.

Enterprise Deployment and Availability

Gemini Cloud Agent is initially targeted at enterprise infrastructure and Google Cloud/Workspace organizations, with broader availability rolling out iteratively across supported platforms and regions.

Teams adopting autonomous coworker agents can monitor execution status, manage API tokens, and configure multi-channel integrations through central enterprise administration dashboards.

Sources

• Original Announcement / Source: FeiZ on X (Post 2108193600815423846)
• Platform Reference: Google Cloud & Gemini Agent Platform