Google’s Gemini Agent Will Route Work to Anthropic’s Claude

Google Cloud announced the Gemini agent on 8 October 2026 at Gemini at Work 2026, describing it as a single, universal agent for work that answers questions, handles knowledge work, creates images and media, and writes and runs code. Google says it runs each job on the model that fits best, orchestrating across its Gemini models and Anthropic’s Claude models today, and other leading private and open models in the future.

In the keynote, adapted on the Google Cloud blog, Thomas Kurian says nearly 500 Google Cloud customers each processed more than one trillion tokens in the last year, and that nearly 90% of the Fortune 100 use Gemini Enterprise.

What the Agent Does

Google says users give the agent objectives, not instructions, and that it plans the work, uses skills and tools, connects to company systems and returns finished work inside documents, inboxes and developer environments.

According to the post, it runs in the cloud with one set of memories, so work that takes hours or days keeps running after a laptop is closed. It keeps four kinds of memory, session, semantic, procedural and episodic, and can create temporary sub-agents for multi-step tasks.

Google also describes coworker agents with a persistent role and their own identity, including their own @agents.company.com email address, which colleagues can add to a chat space or mention in a document.

Where It Works

The post says the agent can be reached from web, mobile and desktop and from the command line, Google Workspace, Microsoft 365 or Slack, and works inline in Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar.

It lists connections to Confluence, Microsoft Office, Teams, Slack, Git, Jira, Salesforce, ServiceNow, BigQuery, Databricks, Postgres and Snowflake, and to any Model Context Protocol server. Industry versions are in preview for financial services and legal, Google says, with government, healthcare and retail coming soon.

Cost and Control

Google writes that per-token prices have dropped 98% since 2024 while enterprise AI volume has exploded, and lists three cost controls: multi-model orchestration, smart routing that sends each workload to the model with the best performance at the lowest cost, and real-time spend caps that pause a project’s agent when its limit is reached.

On governance, it says every agent gets its own cryptographically attested identity, every action is written to an audit trail attributed to the agent, and all traffic passes through an Agent Gateway that enforces company policies.

Google also says its latest TPU 8i system delivers 80% better price-performance than the prior generation.

The Structural Read

The channels reach past Google’s own suite. Google says the agent can be used from Microsoft 365 and Slack as well as Workspace, and connects to any Model Context Protocol server.

Cost control is part of the product. Google lists smart routing, multi-model orchestration and hard per-project spend caps that pause an agent when its budget is reached.

Identity is moving to the agent. Google says each agent gets its own attested identity and audit trail, and coworker agents get their own email address.

Three Implications

MODELS AS A CHOICE Google says the agent orchestrates across Gemini and Claude models today, with other models later.

REACH INTO RIVAL SUITES Google lists Microsoft 365 and Slack among the channels the agent works in.

BUDGETS THAT STOP WORK Real-time spend caps pause a project’s agent at its limit, according to the post.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Agentic Harness War.

The framework’s starting point: “The surface, not the model, governs autonomy.”

As we read it, Google’s launch is that idea stated by a model maker: the post presents the agent, its memory, identity and connectors as the product, and calls the model underneath a separate choice that includes Anthropic’s Claude.

What Is Not Established

We read the Google Cloud blog post and the blog.google announcement. Neither gives pricing for the Gemini agent, a general availability date, or the share of jobs that would run on Claude. Customer results in the post are Google’s accounts of what customers reported. We did not contact Google or Anthropic.

The Bottom Line

Google describes its new Gemini agent as one agent and one API for work, reachable from Workspace, Microsoft 365 and Slack, and says it routes each job across Gemini and Anthropic’s Claude models today, with spend caps and per-agent identity, as nearly 500 of its customers each processed more than a trillion tokens in the last year.