Developer Tools

From developer desks to the whole organization: Running Claude Cowork in Amazon Bedrock

Anthropic's Claude Cowork now runs entirely within AWS Bedrock, keeping all data under enterprise control with no Anthropic licensing.

Deep Dive

Anthropic has announced the availability of Claude Cowork within Amazon Bedrock, marking a significant expansion of its enterprise AI offerings. This integration allows organizations to deploy Claude's desktop application capabilities—including document reading, multi-step research, file processing, and report generation—while keeping all data flows and model inference within their existing AWS environment. Unlike the standard Claude Enterprise offering, Claude Cowork routes all inference exclusively through Amazon Bedrock in the customer's AWS account, ensuring that prompts, files, tool inputs/outputs, and model responses never leave the organization's control and are not used for training foundation models.

The deployment model is designed for enterprise scalability, requiring just two configuration steps: downloading the Claude Desktop application and pushing managed configurations through device management systems like Jamf, Microsoft Intune, or Group Policy. Organizations can leverage existing AWS services for authentication (AWS IAM), network isolation (VPC endpoints), observability (Amazon CloudWatch via OpenTelemetry), and auditing (AWS CloudTrail). Pricing shifts from seat-based licensing to pure consumption-based billing through existing AWS agreements, with granular cost attribution. The solution supports three inference profile types—in-Region, geo cross-Region, and global cross-Region—giving enterprises flexibility in data residency compliance.

For end users, Claude Cowork functions as a powerful desktop agent that can connect to approved external data sources via Model Context Protocol (MCP) servers. This enables practical workflows like a product manager uploading disparate customer notes and requirements for Claude to synthesize into a coherent product brief, complete with technical challenge identification and alternative research. The integration represents a strategic move to transition Claude from a developer-focused tool to an organization-wide AI platform while addressing critical enterprise concerns around security, data sovereignty, and existing cloud infrastructure investment.

Key Points
  • All inference runs through customer's AWS Bedrock account with no data sent to Anthropic for training
  • Consumption-based pricing via AWS billing replaces Anthropic seat licensing, with granular cost attribution
  • Deploys via existing device management (Jamf/Intune) and integrates with AWS IAM, VPC, CloudWatch, and CloudTrail

Why It Matters

Enables secure, scalable deployment of AI agents to all knowledge workers while maintaining data residency and leveraging existing AWS investments.