Quickstart

Welcome to the Amazon Bedrock Agents for Healthcare & Life Sciences (HCLS) contribution guide. This step-by-step tutorial will help you create your own domain-specific agent and contribute it to our growing catalog, as well as configure multiple agents to collaborate effectively.

Prerequisites

Before getting started, ensure you have:

• An AWS account with access to Amazon Bedrock

• Requested access to the foundation models you plan to use with your agent

• AWS CLI configured with appropriate permissions

• Basic knowledge of AWS services and JSON/YAML configuration

• Git and GitHub account (for contributing to the repository)

Creating Your Own Bedrock Agent

• Review Bedrock Agent documentation to build and modify agents
• Review Bedrock Agent documentation to deploy and integrate agents

Using CloudFormation

One-Click Deployment

1. Use the agent example that most closely matches your use case from our catalog
2. Alternatively, create your own using the cdk template located here
3. Wait for the stack deployment to complete (~5-10 minutes)
4. Access your agent through the AWS Bedrock console

Using Jupyter Notebooks

Interactive Development

1. Clone the repository locally:

git clone https://github.com/aws-samples/amazon-bedrock-agents-healthcare-lifesciences.git
cd amazon-bedrock-agents-healthcare-lifesciences

2. Open the agents catalog and navigate to the folder to create your own agent:

cd create-your-own-agent

3. Select from the list of options

   • Use default agent template
   • Bring your own data - SQL database
   • Bring your own data - RAG with documents
   • Bring your own data - Knowledge graph
   • Bring your own tool - lambda function
   • Bring your own tool - custom container
   • Bring your own tool - Foundation model inference with Amazon SageMaker

4. Follow the notebook cells to:

   • Configure your AWS credentials
   • Setup the pre-requisities for data and tools
   • Define your agent’s purpose and behavior
   • Create action groups for API functionality
   • Connect knowledge bases for domain expertise
   • Test and refine your agent’s responses

5. Save your configuration for later contribution. Provide agent description and documentation via Readme.md before submission.

AWS Console

Manual Configuration

1. Log in to the AWS Management Console
2. Navigate to Amazon Bedrock → Agents
3. Click “Create agent”
4. Configure the basic agent settings:
   • Name your agent
   • Select a foundation model (Claude, Titan, etc.)
   • Define instructions for your healthcare/life sciences use case
5. Add Action Groups:
   • Define API schema for domain-specific actions
   • Configure authentication and endpoints
6. Add Knowledge Bases:
   • Connect medical literature or documents
   • Configure retrieval settings
7. Test your agent in the console playground
8. Deploy when satisfied with performance

Agent Configuration Best Practices for HCLS

When creating agents for healthcare and life sciences applications, consider these best practices:

Instructions: Be specific about the HCLS domain expertise your agent should have Context window usage: Structure prompts efficiently to maximize available context Knowledge base optimization: Index documents strategically for relevant retrieval Action group design: Create focused, clear API schemas for data operations Security and compliance: Follow data handling best practices. Use public data sources with appropriate license. Testing: Validate responses with domain-specific test cases

Example agent instructions for a medical terminology agent: You are a healthcare terminology expert specializing in standardized medical coding systems. Your primary expertise includes ICD-10, SNOMED CT, LOINC, RxNorm, and CPT codes.

When responding to queries:

• Provide accurate code lookups when a description is provided
• Explain the meaning of codes when a code is provided
• Suggest related codes when appropriate
• Clarify ambiguities by asking follow-up questions
• Note version differences in coding systems when relevant
• Cite authoritative sources when possible

Do not:

• Provide medical advice or diagnosis
• Make claims about insurance coverage
• Guarantee coding accuracy for billing purposes
• Share information about specific patients

Use your knowledge bases to reference official coding manuals and guidelines. Use API actions to look up specific codes when needed.

Contributing Your Agent to the Catalog

Once you’ve created and tested your agent, follow these steps to contribute to our catalog:

1. Fork the repository on GitHub

2. Clone your fork locally:

 git clone https://github.com/YOUR-USERNAME/amazon-bedrock-agents-healthcare-lifesciences.git
 cd amazon-bedrock-agents-healthcare-lifesciences

 #Create a new branch:

 git checkout -b add-my-agent-name

3. Add your agent to the catalog:

• Create a new directory in agents_catalog/ with your agent’s name
• Include the following files:
  • README.md: Description, use cases, and setup instructions
  • cloudformation/*.yaml: Templates for one-click deployment
  • notebooks/: Jupyter notebooks for custom setup and test cases
  • sample-prompts.md: Example interactions showing capabilities
  • architecture.png: Visual diagram of your agent

Submit a pull request: Go to the original repository Click “Pull Requests” → “New Pull Request” Select your fork and branch Complete the PR template with details about your agent

Our team will review your submission and provide feedback before merging.

Configuring Multiple Agents in Collaboration

Orchestration Setup

The Bedrock multi-agent collaboration framework allows multiple specialized agents to collaborate on complex queries.

• Review Bedrock Agent documentation to use multi-agent collaboration

Using Runtime initialization

Select agents and configure automatically

1. Launch the Configure Multi-agent collaboration UI
2. Select agents and hit configure
3. Test orchestration with your queries

Using Jupyter Notebooks

Interactive Development

1. Clone the repository branch with multi-agent features:

   git clone https://github.com/aws-samples/amazon-bedrock-agents-healthcare-lifesciences.git
   cd amazon-bedrock-agents-healthcare-lifesciences/multi_agent_collaboration

2. Review the directory structure to understand the components:

   multi_agent_collaboration/
   ├── templates/            # CloudFormation templates
   ├── examples/             # Example implementations
   └── README.md             # Documentation

3. Follow the notebook cells in the example notebook to:

   • Configure your AWS credentials
   • Setup the pre-requisities for sub-agents
   • Define your supervisor agent’s purpose and behavior
   • Select Supervisor mode or Supervisor with router mode
   • Associate the list of agents
   • Test and refine your supervisor agent’s responses

4. Save your configuration for later contribution. Provide agent description and documentation via Readme.md before submission.

AWS Console

Manual Configuration

1. Log in to the AWS Management Console
2. Navigate to Amazon Bedrock → Agents
3. Create collaborator agents
4. Create a new supervisor agent or assign supervisor role to an existing agent
   • Choose between Supervisor or Supervsior with routing mode
5. Add collaborator agents
6. Test your agent in the console playground
7. Deploy when satisfied with performance

Conclusion

Congratulations! You’ve learned how to:

• Create a specialized Bedrock agent for healthcare and life sciences
• Contribute your agent to the community catalog
• Configure multiple agents to collaborate on complex queries
• Deploy, test, and monitor your multi-agent system

By following this guide, you’ve added valuable capabilities to the AWS Bedrock Agents for Healthcare & Life Sciences ecosystem. Your contribution will help healthcare organizations leverage AI to improve patient care, streamline operations, and accelerate research.

Next Steps

Explore advanced agent customization options in the AWS Bedrock documentation