In this article
  1. What happened
  2. Why it matters
  3. What to do now
  4. What’s still unknown

TL;DR: On October 22, 2026, Anthropic and Google Cloud made “Claude Models on Google Cloud (Anthropic managed)” generally available. It gives Google Cloud customers the full Claude Platform, including Claude Managed Agents (beta), skills, code execution, computer use, the Files API and the Claude Console, with Google Cloud sign-in (IAM), one Google Cloud bill, and usage that counts toward eligible Google Cloud spend commitments. The trade-off is that Anthropic, not Google, is the data processor. The existing Google-managed option stays for teams that need Google as processor, regional data residency, or tools such as BigQuery and Model Armor. Google says pricing is the same on both. If your business already pays Google Cloud and wants Claude’s newest agent features, this is now the simplest route.

What happened

Anthropic and Google Cloud announced the launch on October 22. Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5 are available, and Anthropic says new models will arrive on this option as they launch. According to the two announcements and Anthropic’s setup documentation:

  • Features on day one. New Claude Platform features and betas ship to the Anthropic-managed option the same day they go live on Anthropic’s own API. That includes Claude Managed Agents (beta), code execution, computer use, Agent Skills, the Files API, citations, web search and fetch, the MCP connector (beta), prompt caching and batch processing. Fast mode for Claude Opus 5.5 is also generally available here at $8 per million input tokens and $40 per million output tokens.
  • Google Cloud sign-in and billing. Access runs through Google Cloud IAM, with three roles: Claude Viewer (read-only), Claude User (can send requests and manage files, skills, batches and agents) and Claude Admin (organization-level settings, and the only role that can open the Claude Console). Anthropic meters usage and reports it to Google, and the charges appear on your Google Cloud bill with no separate Anthropic invoice.
  • Who processes your data. Anthropic is the data processor under its Commercial Terms, and some supporting services, such as its trust and safety systems, process data outside Google Cloud. The service follows the Claude API’s data retention policy, and Zero Data Retention is available on request.
  • Endpoints. The service is available with US and Global endpoints.

The older Google-managed option continues. Google remains the data processor there, and it offers VPC Service Controls, customer-managed encryption keys, data residency in the US, EU and APAC, and native links to BigQuery and Model Armor. (Anthropic’s docs say customer-managed encryption keys with Cloud KMS are also available on the Anthropic-managed option.) New Claude Platform features reach it later, “once integrated,” Google says. Google’s comparison table lists the same models on day zero and standard pricing for both options.

Why it matters

Until now, Google Cloud customers who used Claude got the models but not always Anthropic’s newest tools, which arrive first on Anthropic’s own API. That mattered most for agent work: things like Managed Agents, skills and the MCP connector are what turn a chat model into something that runs a task end to end.

For a small business or startup that already runs on Google Cloud, the practical gains are:

  1. One bill and existing credits or commitments. Claude usage counts toward a Google Cloud spend commitment the way other Marketplace purchases do, so you don’t need a separate Anthropic contract or card on file.
  2. No API keys to leak. This option doesn’t use API keys at all. Calls authenticate with Google Cloud credentials, which removes a common way secrets end up in code or logs.
  3. Near-full feature parity. Anthropic’s docs say three things differ from a regular Claude API request: the base URL, Google Cloud authentication in place of an API key, and the workspace ID, which moves into the URL. Request bodies, streaming and responses stay the same. A few features aren’t supported, including MCP tunnels, memory stores on self-hosted environments and the pre-built self-hosted sandbox workers. If you’re also updating your SDKs, check our note on the Claude API version 2026-10-21 changes first.

The catch is that setup creates a new Anthropic organization tied to your Google Cloud project. It’s separate from any Anthropic account your company already has, including Claude Enterprise. API keys, workspaces, Console settings, files, skills and agent resources don’t carry over, and a Zero Data Retention arrangement on another Anthropic organization doesn’t apply. Anthropic also says that if you have an existing discount, you should contact your Anthropic or Google Cloud account representative before you start, because discounts can’t be applied to past usage.

This follows Google Cloud’s October 21 announcement that OpenAI models are coming to Google Cloud in private preview. Both announcements let Google Cloud customers pay for another AI lab’s models through their Google Cloud bill.

What to do now

  1. Decide which option you need. Choose Anthropic managed if you want new Claude features right away. Stay on Google managed if a customer contract or regulation requires Google as processor, regional data residency, or VPC Service Controls.
  2. Ask about discounts first. If you already have a discount with Anthropic or Google Cloud, contact your account representative before subscribing.
  3. Follow the setup steps in order. Subscribe to the “Claude Models on Google Cloud (Anthropic managed)” listing in Google Cloud Marketplace. Enable claude.googleapis.com and anthropic-managed-claude.cloudpartnerservices.goog on each project that will call Claude. Activate the Anthropic organization, then copy the workspace ID from the Claude Console for your developers.
  4. Give least access. Grant roles/claude.user to the service account that calls Claude, and keep roles/claude.admin to the few people who manage settings and billing.
  5. Plan for agent sessions. Anthropic’s docs warn that revoking someone’s Google Cloud access doesn’t stop Managed Agents sessions or scheduled deployments they already created. When someone leaves or you rotate a service account, review and stop those sessions yourself.
  6. Turn on Data Access audit logs for claude.googleapis.com if you want a record of requests. They’re off by default.

What’s still unknown

Anthropic says rate limits are managed by Anthropic and higher limits go through its sales team, but it hasn’t published default limits for this option. The SDK clients for the Anthropic-managed option are in beta. Neither company has said when the Google-managed option will get specific newer features, such as Managed Agents.

Sources & method

Sources

Method. Last checked October 23, 2026, at 7:30 a.m. Manila time (23:30 UTC on October 22). Features, models, roles, setup steps, billing, data processing and limits come from Anthropic’s announcement and documentation. The comparison of the two options and the pricing statement come from Google Cloud’s blog. The advice on choosing an option, access and session clean-up is our own judgment, and we haven’t set up the service ourselves. No affiliate links on this page. How we work.

Spotted an error? See our corrections policy and tell us.