Google Cloud unveils Gemini agent that routes work to Claude

Google Cloud said Thursday in a blog post by Chief Executive Thomas Kurian that it unveiled Gemini as a single work agent and a single API for questions, knowledge work, media and code. The agent is in private preview, 9to5Google reported.

How it works: Staff give the agent an objective rather than a step list, Kurian wrote. It keeps session, semantic, procedural and episodic memory, can spin up temporary sub-agents for multi-step jobs, and can sit as a coworker with its own email and storage. Every action is logged to the agent rather than a person, Google said.

Which models: Google says the agent runs each job on the Gemini family or on Anthropic’s Claude today, with other private and open models later, plus real-time spend caps and smart routing. Connectors named in the post include Workspace, Microsoft Office, Teams, Slack, Jira, Confluence, Git, Salesforce, ServiceNow, BigQuery, Databricks, Postgres, Snowflake, desktop files and any Model Context Protocol server. It can be called from the web, a phone, a desktop, the command line, or run headless inside another app. Work that takes hours or days keeps going after the laptop closes, Google says.

Inside Workspace: Google says Gemini works inline in Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar.

Who is using it: Kurian said nearly 500 Google Cloud customers each processed more than 1 trillion tokens in the last year, nearly 80% of Google Cloud customers use Google AI products and nearly 90% of the Fortune 100 use Gemini Enterprise. On tested the model-selection feature, Google said, and it cited Shopify and PayPal as existing multi-model users. Google said Orange Spain is deploying more than 1,000 custom Gemini Enterprise agents.

What’s next: Wider availability for select Workspace Business and Enterprise plans is coming soon, with no date, 9to5Google reported. Versions for government, health care and retail are “coming soon,” Google said.

— Rowan Keel