Claude Cowork is Now Available on Amazon Bedrock

Anthropic’s Claude Cowork is now available on Amazon Bedrock. While Claude Code for developers has been gaining attention, this announcement enables knowledge workers across the organization to leverage AI.

Claude Cowork is a desktop application that automates document reading, multi-step research, file processing, and report generation. It includes all the features of the traditional Claude Desktop (projects, artifacts, memory, file upload, remote connectors, skills, plugins, and MCP servers).

A key point is that the inference processing is done entirely through Amazon Bedrock. This means that data remains within the organization’s AWS account, and Amazon Bedrock does not store prompts, files, tool inputs/outputs, or model responses, nor does it use them for base model training. (Source: aws.amazon.com)

Setup Procedure and Technical Requirements

The integration of Claude Cowork with Amazon Bedrock is completed in two steps.

First, users download the Claude Desktop application to their machines. Then, a device management system (such as Jamf, Microsoft Intune, or Group Policy) delivers settings to Claude Desktop and enables inference mode.

The setup requires specifying the following:

  • Model ID and Amazon Bedrock Inference Profile
  • Authentication method
  • Organizational policies

For organizations using the LLM Gateway, the same management settings can be used by specifying the gateway URL. If you already have Claude Code running on Amazon Bedrock, you can reuse the same setup.

# Example of checking Bedrock model access using AWS CLI
aws bedrock list-foundation-models --region us-east-1 \
  --by-provider anthropic

For organization-wide deployment, access to Bedrock in the AWS account, IAM permissions, explicit region and network settings, and operational guardrails are necessary. (Source: elevata.io)

Performance Optimization with Prompt Caching

The combination of Amazon Bedrock’s prompt caching feature with Claude Code dramatically improves AI-assisted development for large codebases.

Prompt caching allows Amazon Bedrock to interpret cache checkpoint markers defined by the application and save the entire model state after text processing. Subsequent requests that reuse the same prefix load the cached state instead of recalculating.

Claude Code intelligently manages these cache points during codebase processing. When asking multiple questions about the same code or iteratively refining solutions, Claude Code can “remember” previously analyzed code without incurring full computational and financial costs.

# Example of using prompt caching with Bedrock Converse API
import boto3

bedrock = boto3.client('bedrock-runtime', region_name='us-east-1')

response = bedrock.converse(
    modelId='anthropic.claude-3-5-sonnet-20241022-v2:0',
    messages=[{
        'role': 'user',
        'content': [{'text': 'Analyze this codebase...'}]
    }],
    inferenceConfig={
        'maxTokens': 4096,
        'temperature': 0.1
    }
)

This combination results in reduced response times and input token costs, making AI-assisted coding for daily development tasks more efficient and economical. (Source: aws.amazon.com)

Enterprise Agent AI Platform

The combination of Amazon Bedrock AgentCore and Claude is transforming enterprise-level agent AI deployments. At Cox Automotive, 17 major proof-of-concepts have been deployed to production, and 7 industry-transforming solutions are currently under development.

Traditionally, when enterprises built agent AI, engineering teams spent months building infrastructure for session management, authentication bolts, tool orchestration, observability frameworks, and scaling logic. By the time they could focus on actual agent logic and business value, teams were exhausted, and use cases had evolved.

Amazon Bedrock AgentCore is a comprehensive agent platform for building, deploying, and operating high-performance agents at scale. It is model-agnostic and handles infrastructure and operational challenges, allowing developers to focus on business-differentiating elements (agent logic and specific tasks to execute).

The fully managed AgentCore service includes Runtime, Memory, Identity, Gateway, Code Interpreter, Browser Tool, and Observability, providing up to 8 hours of autonomous execution and automatic scaling. (Source: aws.amazon.com)

Summary

  • Introducing Claude Cowork on Amazon Bedrock enables knowledge workers across the organization, not just developers, to leverage AI for document processing and research automation
  • Existing Claude Code setups can be reused, and device management systems can deliver settings for rapid enterprise-wide deployment
  • The prompt caching feature significantly reduces response times and costs for iterative AI-assisted development on large codebases
  • Utilizing Amazon Bedrock AgentCore’s fully managed infrastructure allows engineers to focus on agent business logic instead of infrastructure building, enabling the creation of autonomous agents that can run for up to 8 hours