Anthropic has opened its own managed version of Claude to Google Cloud customers, with the complete Claude Platform behind it. The company announced general availability of “Claude Models on Google Cloud (Anthropic managed)” in a post on claude.com on 22 October 2026. Anthropic runs this service itself, and that makes Anthropic the data processor.
Sign-in and billing stay on the Google side. Teams use the Google Cloud IAM credentials they already manage, so existing permissions carry over. Charges land on one invoice from Google, and Claude consumption counts against whatever spend commitment the customer has already signed with Google. For a finance team, that means Claude can be paid for from a cloud budget that is already approved, with no new vendor contract to negotiate.
The feature list is the point of the launch. Anthropic says new features and betas reach this offering on their launch day, matching its native platform. Customers get:
- Claude Managed Agents (beta), for long-running agents without managing infrastructure
- Code execution, computer use, and agent skills
- An advisor strategy (beta), where an agent consults a second model
- The Files API, citations, web search, and web fetch
- An MCP connector (beta), which links Claude to other tools without client code
- Prompt caching and batch processing for cost and volume
Customers also get the Claude Console, Anthropic’s workspace for writing prompts, configuring agents, and watching usage and costs. The offering currently has US and Global endpoints. Four models are offered: Haiku 5.5, Sonnet 5.5, Opus 5.5, and Fable 5.1. Anthropic says new models will arrive as they launch.
Opus 5.5 also has Fast Mode in general availability on this route. Anthropic says it produces output up to 2.5 times faster, priced at $8 per million input tokens and $40 per million output tokens.
The older route is not going away. Anthropic describes the Google managed offering as the better fit for customers who require Google as their data processor, face tight regional data residency rules, or want native Google Cloud pieces such as Model Armor or Agent Engine. So the choice is a trade: full and immediate feature access with Anthropic as processor, or Google as processor and the native Google integrations. The post does not say whether per-token prices match between the two.
The timing is notable. One day earlier, OpenAI said its models are coming to Google Cloud as well, which would put two rival labs on the same cloud shelf and leave the buyer to choose on features and data terms rather than on which cloud they already use.
One warning sits at the end of Anthropic’s post. Customers who hold an existing discount should talk to their Anthropic or Google account rep before they start, because the discount will not reach usage that has already happened. Any team planning to move Claude traffic over this quarter should settle that conversation before the first request goes out.
Reported by Anthropic on claude.com, 22 October 2026.
