Google Cloud launches Gemini agent: universal work agent with Gemini+Claude routing, sub-agents and coworker identities, multi-day cloud persistence

At Gemini at Work 2026 (Oct 8), Google Cloud CEO Thomas Kurian introduced the Gemini agent: a single universal work agent for Q&A, knowledge work, media creation, and writing/running code from one prompt box—you give objectives, not step lists. It runs persistently in the cloud (hours/days with laptop closed), spins temporary sub-agents and coworker agents with their own identity and @agents.company.com email, and keeps four memory types (session, semantic, procedural, episodic). Tools/skills registries connect Workspace, Microsoft 365, Slack, Git, Jira, Salesforce, BigQuery, MCP servers, and more. Model choice is decoupled from the agent: jobs can route across Gemini and Anthropic Claude today, with more private/open models later. Governance includes Agent Identity, Agent Gateway, sandboxes, Smart Routing, and per-project spend caps. Inline Workspace (@Gemini) plus Workspace Intelligence for one-click deferrable tasks. No new public coding benchmarks; customer ops anecdotes are not reproducible leaderboards. Availability: private preview now, broader roll out for select Workspace Business/Enterprise. Primary source: https://cloud.google.com/blog/products/ai-machine-learning/welcome-to-gemini-at-work-2026