Supercharge your improvement with Claude Code and Amazon Bedrock immediate caching

Immediate caching in Amazon Bedrock is now usually obtainable, delivering efficiency and price advantages for agentic AI functions. Coding assistants that course of giant codebases signify a super use case for immediate caching.

On this publish, we’ll discover find out how to mix Amazon Bedrock immediate caching with Claude Code—a coding agent launched by Anthropic that’s now usually obtainable. This highly effective mixture transforms your improvement workflow by delivering lightning-fast responses from decreasing inference response latency, in addition to reducing enter token prices. You’ll uncover how this makes AI-assisted coding not simply extra environment friendly, but in addition extra economically viable for on a regular basis improvement duties.

What’s Claude Code?

Claude Code is Anthropic’s AI coding assistant powered by Claude Sonnet 4. It operates instantly in your terminal, your favourite IDEs reminiscent of VS Code and Jetbrains, and within the background with Claude Code SDK, understanding your challenge context and taking actions with out requiring you to manually manipulate and add generated code to a challenge. Not like conventional coding assistants, Claude Code can:

• Write code and repair bugs spanning a number of information throughout your codebase
•  Reply questions on your code’s structure and logic
• Execute and repair exams, linting, and different instructions
• Search by git historical past, resolve merge conflicts, and create commits and PRs
• Function all your different command line instruments, like AWS CLI, Terraform, and k8s

Essentially the most compelling facet of Claude Code is the way it integrates into your present workflow. You merely level it to your challenge listing and work together with it utilizing pure language instructions. Claude Code additionally helps Mannequin Context Protocol (MCP), permitting you to attach exterior instruments and knowledge sources on to your terminal and customise its AI capabilities together with your context.

To study extra, see Claude Code tutorials and Claude Code: Finest practices for agentic coding.

Amazon Bedrock immediate caching for AI-assisted improvement

The immediate caching function of Amazon Bedrock dramatically reduces each response occasions and prices when working with giant context. Right here’s the way it works: When immediate caching is enabled, your agentic AI utility (reminiscent of Claude Code) inserts cache checkpoint markers at particular factors in your prompts. Amazon Bedrock then interprets these application-defined markers and creates cache checkpoints that save the complete mannequin state after processing the previous textual content. On subsequent requests, in case your immediate reuses that very same prefix, the mannequin masses the cached state as an alternative of recomputing.

Within the context of Claude Code particularly, this implies the appliance intelligently manages these cache factors when processing your codebase, permitting Claude to “bear in mind” beforehand analyzed code with out incurring the complete computational and monetary price of reprocessing it. While you ask a number of questions on the identical code or iteratively refine options, Claude Code leverages these cache checkpoints to ship quicker responses whereas dramatically decreasing token consumption and related prices.

To study extra, see documentation for Amazon Bedrock immediate caching.

Answer overview: Strive Claude Code with Amazon Bedrock immediate caching

Conditions

Immediate caching is routinely turned on for supported fashions and AWS Areas.

Establishing Claude Code with Claude Sonnet 4 on Amazon Bedrock

After configuring AWS CLI together with your credentials, observe these steps:

1. In your terminal, execute the next instructions:

   # Set up Claude Code
   npm set up -g @anthropic-ai/claude-code
   
   # Configure for Amazon Bedrock
   export CLAUDE_CODE_USE_BEDROCK=1
   export ANTHROPIC_MODEL='us.anthropic.claude-sonnet-4-20250514-v1:0'
   export ANTHROPIC_SMALL_FAST_MODEL='us.anthropic.claude-3-5-haiku-20241022-v1:0'
   
   # Launch Claude Code
   claude

2. Confirm that Claude Code is working by checking for the Welcome to Claude Code! message in your terminal.
   

To study extra about find out how to configure Claude Code for Amazon Bedrock, see Connect with Amazon Bedrock.

Getting began with immediate caching

To get began, let’s experiment with a easy immediate.

1. In Claude Code, execute the immediate:

   construct a fundamental text-based calculator

2. Evaluate and reply to Claude Code’s requests:
   1. When prompted with questions like Do you need to create calculator.py? choose 1. Sure to proceed.
      Instance query:

      Do you need to create calculator.py?
      
      1. Sure
      2. Sure, and do not ask once more for this session (shift+tab)
      3. No, and inform Claude what to do in another way (esc)

   2. Fastidiously overview every request earlier than approving to keep up safety.
3. After Claude Code generates the calculator utility, it’s going to show execution directions reminiscent of:

   Run the calculator with: python3 calculator.py

4. Take a look at the appliance by executing the instructed command above. Then, observe the on-screen prompts to carry out calculations.

Claude Code routinely allows immediate caching to optimize efficiency and prices. To watch token utilization and prices, use the /price command. You’ll obtain an in depth breakdown just like this:

/price 
  ⎿  Complete price:            $0.0827
  ⎿  Complete length (API):  26.3s
  ⎿  Complete length (wall): 42.3s
  ⎿  Complete code adjustments:    62 strains added, 0 strains eliminated

This output gives helpful insights into your session’s useful resource consumption, together with whole price, API processing time, wall clock time, and code modifications.

Getting began with immediate caching

To know the advantages of immediate caching, let’s strive the identical immediate with out immediate caching for comparability:

1. Within the terminal, exit Claude Code by urgent Ctrl+C.
2. To create a brand new challenge listing, run the command:

   mkdir test-disable-prompt-caching; cd test-disable-prompt-caching 

3. Disable immediate caching by setting an atmosphere variable:

   export DISABLE_PROMPT_CACHING=1

4. Execute claude to run Claude Code.
5. Confirm immediate caching is disabled by checking the terminal output. It’s best to see Immediate caching: off underneath the Overrides (through env) part.
   
6. Execute the immediate:

   construct a fundamental text-based calculator

7. After completion, execute /price to view useful resource utilization.

You will note the next useful resource consumption in comparison with when immediate caching is enabled, even with a easy immediate:

/price 
  ⎿  Complete price:            $0.1029
  ⎿  Complete length (API):  32s
  ⎿  Complete length (wall): 1m 17.5s
  ⎿  Complete code adjustments:    57 strains added, 0 strains eliminated

With out immediate caching, every interplay incurs the complete price of processing your context.

Cleanup

To re-enable immediate caching, exit Claude Code and run unset DISABLE_PROMPT_CACHING earlier than restarting Claude. Claude Code doesn’t incur price when you find yourself not utilizing it.

Immediate caching for complicated codebases and environment friendly iteration

When working with complicated codebases, immediate caching delivers considerably higher advantages than with easy prompts. For an illustrative instance, take into account the preliminary immediate: Develop a recreation just like Pac-Man. This preliminary immediate generates the foundational challenge construction and information. As you refine the appliance with prompts reminiscent of Implement distinctive chase patterns for various ghosts, the coding agent should comprehend your whole codebase to have the ability to make focused adjustments.

With out immediate caching, you drive the mannequin to reprocess 1000’s of tokens representing your code construction, class relationships, and present implementations, with every iteration.

Immediate caching alleviates this redundancy by preserving your complicated context, reworking your software program improvement workflow with:

• Dramatically diminished token prices for repeated interactions with the identical information
• Sooner response occasions as Claude Code doesn’t must reprocess your whole codebase
• Environment friendly improvement cycles as you iterate with out incurring full prices every time

Immediate caching with Mannequin Context Protocol (MCP)

Mannequin Context Protocol (MCP) transforms your coding expertise by connecting coding brokers to your particular instruments and data sources. You’ll be able to join Claude Code to MCP servers that combine to your file methods, databases, improvement instruments and different productiveness instruments. This transforms a generic coding assistant into a personalised assistant that may work together together with your knowledge and instruments past your codebase, observe your group’s greatest practices, accelerating your distinctive improvement processes and workflows.

While you construct on AWS, you acquire further benefits by leveraging AWS open supply MCP servers for code assistants that present clever AWS documentation search, best-practice suggestions, and real-time price visibility, evaluation and insights – with out leaving your software program improvement workflow.

Amazon Bedrock immediate caching turns into important when working with MCP, because it preserves complicated context throughout a number of interactions. With MCP repeatedly enriching your prompts with exterior data and instruments, immediate caching alleviates the necessity to repeatedly course of this expanded context, slashing prices by as much as 90% and decreasing latency by as much as 85%. This optimization proves notably helpful as your MCP servers ship more and more refined context about your distinctive improvement atmosphere, so you may quickly iterate by complicated coding challenges whereas sustaining related context for as much as 5 minutes with out efficiency penalties or further prices.

Concerns when deploying Claude Code to your group

With Claude Code now usually obtainable, many shoppers are contemplating deployment choices on AWS to reap the benefits of its coding capabilities. For deployments, take into account your foundational structure for safety and governance:

Take into account leveraging AWS IAM Id Middle, previously AWS Single Signal On (SSO) to centrally govern id and entry to Claude Code. This verifies that solely licensed builders have entry. Moreover, it permits builders to entry assets with non permanent, role-based credentials, assuaging the necessity for static entry keys and enhancing safety. Previous to opening Claude Code, just be sure you configure AWS CLI to make use of an IAM Id Middle profile by utilizing aws configure sso --profile . Then, you login utilizing the profile created aws sso login --profile .

Take into account implementing a generative AI gateway on AWS to trace and attribute prices successfully throughout completely different groups or tasks utilizing inference profiles. For Claude Code to make use of a customized endpoint, configure the ANTHROPIC_BEDROCK_BASE_URL atmosphere variable with the gateway endpoint. Be aware that the gateway ought to be a pass-through proxy, see instance implementation with LiteLLM. To study extra about AI gateway options, contact your AWS account workforce.

Take into account automated configuration of default atmosphere variables. This consists of the atmosphere variables outlined on this publish, reminiscent of CLAUDE_CODE_USE_BEDROCK, ANTHROPIC_MODEL, and ANTHROPIC_FAST_MODEL. It will configure Claude Code to routinely join Bedrock, offering a constant baseline for improvement throughout groups. To start with, organizations can begin by offering builders with self-service directions.

Take into account permissions, reminiscence and MCP servers in your group. Safety groups can configure managed permissions for what Claude Code is and isn’t allowed to do, which can’t be overwritten by native configuration. As well as, you may configure reminiscence throughout all tasks which lets you auto-add frequent bash instructions information workflows, and magnificence conventions to align together with your group’s desire. This may be executed by deploying your CLAUDE.md file into an enterprise listing //CLAUDE.md or the person’s residence listing ~/.claude/CLAUDE.md. Lastly, we advocate that one central workforce configures MCP servers and checks a .mcp.json configuration into the codebase so that every one customers profit.

To study extra, see Claude Code workforce setup documentation or contact your AWS account workforce.

Conclusion

On this publish, you discovered how Amazon Bedrock immediate caching can considerably improve AI functions, with Claude Code’s agentic AI assistant serving as a strong demonstration. By leveraging immediate caching, you may course of giant codebases extra effectively, serving to to dramatically scale back prices and response occasions. With this expertise you may have quicker, extra pure interactions together with your code, permitting you to iterate quickly with generative AI. You additionally discovered about Mannequin Context Protocol (MCP), and the way the seamless integration of exterior instruments enables you to customise your AI assistant with particular context like documentation and net assets. Whether or not you’re tackling complicated debugging, refactoring legacy methods, or creating new options, the mix of Amazon Bedrock’s immediate caching and AI coding brokers like Claude Code affords a extra responsive, cost-effective, and clever method to software program improvement.

Amazon Bedrock immediate caching is mostly obtainable with Claude 4 Sonnet and Claude 3.5 Haiku. To study extra, see immediate caching and Amazon Bedrock.

Anthropic Claude Code is now usually obtainable. To study extra, see Claude Code overview and phone your AWS account workforce for steerage on deployment.

Concerning the Authors

Jonathan Evans is a Worldwide Options Architect for Generative AI at AWS, the place he helps prospects leverage cutting-edge AI applied sciences with Anthropic’s Claude fashions on Amazon Bedrock, to resolve complicated enterprise challenges. With a background in AI/ML engineering and hands-on expertise supporting machine studying workflows within the cloud, Jonathan is keen about making superior AI accessible and impactful for organizations of all sizes.

Daniel Wirjo is a Options Architect at AWS, targeted on SaaS and AI startups. As a former startup CTO, he enjoys collaborating with founders and engineering leaders to drive development and innovation on AWS. Exterior of labor, Daniel enjoys taking walks with a espresso in hand, appreciating nature, and studying new concepts.

Omar Elkharbotly is a Senior Cloud Help Engineer at AWS, specializing in Information, Machine Studying, and Generative AI options. With intensive expertise in serving to prospects architect and optimize their cloud-based AI/ML/GenAI workloads, Omar works carefully with AWS prospects to resolve complicated technical challenges and implement greatest practices throughout the AWS AI/ML/GenAI service portfolio. He’s keen about serving to organizations leverage the complete potential of cloud computing to drive innovation in generative AI and machine studying.

Gideon Teo is a FSI Answer Architect at AWS in Melbourne, the place he brings specialised experience in Amazon SageMaker and Amazon Bedrock. With a deep ardour for each conventional AI/ML methodologies and the rising area of Generative AI, he helps monetary establishments leverage cutting-edge applied sciences to resolve complicated enterprise challenges. Exterior of labor, he cherishes high quality time with family and friends, and repeatedly expands his data throughout numerous expertise domains.