Anthropic Launches Managed Claude on Google Cloud

Anthropic said it officially launched Managed Claude on Google Cloud on October 22 local time, giving customers access to all Claude Platform features through Google Cloud. Customers can use Google Cloud account and permissions management and billing while accessing Claude features and beta services operated directly by Anthropic.

The service supports existing Google Cloud IAM credentials and connects through application default credentials without a separate API key. Charges are added to Google Cloud bills, and customers can apply credits under existing Google Cloud commitments. New Claude Platform features and betas become available in Managed Claude on the same day. The service is available through US and global endpoints.

Developers can use the Claude Managed Agents beta to build long-running AI agents, along with code execution, computer use and agent skills. Code execution runs Python within API calls for visualization and data analysis; computer use operates browsers or desktops to complete tasks. The Files API lets users upload and reference documents across conversations, while citations provide sources for answers. Web search and web page retrieval provide information, and the MCP connector beta connects to external tools without requiring developers to write separate client code.

Prompt caching stores repeated context to reduce costs and latency, while batch processing handles large workloads asynchronously. In the Claude Console, users can write prompts, configure and run agents, monitor usage and costs, and manage workspace access permissions and security. Fast mode for Claude Opus 5.5 is also generally available. It prioritizes speed and raises output token generation speed by up to 2.5x. Anthropic said the feature suits real-time debugging, rapid iterative development and conversational agents. Pricing is $8 per million input tokens and $40 per million output tokens.

Four models are currently available: Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5. Anthropic plans to add future models to the managed service when they launch. Google Cloud customers can choose between Anthropic-managed and existing Google-managed options. Anthropic-managed service is intended for businesses seeking quick access to the latest features and betas. Although it runs on Google Cloud infrastructure, Anthropic retains operational, legal and data-processing responsibility under its commercial terms.

Under the existing Google-managed option, Google acts as the data processor. That option suits companies with strict data residency requirements or those seeking close integration with Google Cloud services such as Model Armor and Agent Engine. Managed Claude is available through Google Cloud Marketplace. Customers with existing discount agreements should confirm with their Anthropic or Google Cloud representative whether the discount applies before use; Google Cloud commitments and discounts cannot be applied retroactively to charges already incurred.

Anthropic said it officially launched Managed Claude on Google Cloud on October 22 local time, giving customers access to all Claude Platform features through Google Cloud. Customers can use Google Cloud account and permissions management and billing while accessing Claude features and beta services operated directly by Anthropic.

The service supports existing Google Cloud IAM credentials and connects through application default credentials without a separate API key. Charges are added to Google Cloud bills, and customers can apply credits under existing Google Cloud commitments. New Claude Platform features and betas become available in Managed Claude on the same day. The service is available through US and global endpoints.

Developers can use the Claude Managed Agents beta to build long-running AI agents, along with code execution, computer use and agent skills. Code execution runs Python within API calls for visualization and data analysis; computer use operates browsers or desktops to complete tasks. The Files API lets users upload and reference documents across conversations, while citations provide sources for answers. Web search and web page retrieval provide information, and the MCP connector beta connects to external tools without requiring developers to write separate client code.

Prompt caching stores repeated context to reduce costs and latency, while batch processing handles large workloads asynchronously. In the Claude Console, users can write prompts, configure and run agents, monitor usage and costs, and manage workspace access permissions and security. Fast mode for Claude Opus 5.5 is also generally available. It prioritizes speed and raises output token generation speed by up to 2.5x. Anthropic said the feature suits real-time debugging, rapid iterative development and conversational agents. Pricing is $8 per million input tokens and $40 per million output tokens.

Four models are currently available: Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5. Anthropic plans to add future models to the managed service when they launch. Google Cloud customers can choose between Anthropic-managed and existing Google-managed options. Anthropic-managed service is intended for businesses seeking quick access to the latest features and betas. Although it runs on Google Cloud infrastructure, Anthropic retains operational, legal and data-processing responsibility under its commercial terms.

Under the existing Google-managed option, Google acts as the data processor. That option suits companies with strict data residency requirements or those seeking close integration with Google Cloud services such as Model Armor and Agent Engine. Managed Claude is available through Google Cloud Marketplace. Customers with existing discount agreements should confirm with their Anthropic or Google Cloud representative whether the discount applies before use; Google Cloud commitments and discounts cannot be applied retroactively to charges already incurred.