This toolkit provides a library of reusable AI capabilities, skills, and tools specifically for the Healthcare and Life Sciences (HCLS) domain on AWS. It enables developers, researchers, and platform teams to build and deploy sophisticated AI agents and integrate HCLS domain knowledge into various AI platforms, including coding assistants and end-user applications, accelerating innovation in areas like drug discovery and genomics.

How It Works

The core approach involves decomposing over 36 reference agents into portable "skills" (domain knowledge) and "MCP servers" (domain tools). This modular design allows HCLS capabilities to be consumed independently across different platforms—such as AI coding assistants (Claude Code, Kiro) and end-user applications (Amazon Quick)—or deployed as full agents on Amazon Bedrock AgentCore. This "portable interface layer" abstracts domain logic, enabling natural language interactions with complex HCLS data and workflows without requiring full agent deployment for every use case.

Quick Start & Requirements

• Primary install: Clone the repository (git clone https://github.com/aws-samples/amazon-bedrock-agents-healthcare-lifesciences.git). Follow platform-specific setup instructions found in the platforms/ directory (e.g., platforms/claude-code/README.md). For AgentCore deployments, use agentcore_template/ and install dependencies via uv pip install -r dev-requirements.txt.
• Prerequisites: AWS Agent Toolkit (for generic infrastructure skills), Python 3.x, virtual environment (.venv), uv pip. Specific platforms may have additional requirements.
• Links:
  • Claude Code Setup: platforms/claude-code/README.md
  • Kiro Setup: platforms/kiro/README.md
  • Amazon Quick Setup: platforms/amazon-quick/
  • AgentCore Template: agentcore_template/
  • Kiro Power for Agent Development: powers/hcls-agentcore-builder/

Highlighted Details

• Features a catalog of 36+ specialized reference agents for drug research, clinical trials, genomics, and commercialization.
• Integrates with key HCLS data sources and ontologies, including Biomni (30+ databases), HealthOmics, OLS (200+ ontologies), PubMed, and Open Targets.
• Provides portable skills and MCP servers designed for seamless integration with AI coding assistants (Claude Code, Kiro) and end-user platforms (Amazon Quick).
• Includes a Kiro Power (powers/hcls-agentcore-builder/) to assist developers in building AgentCore agents.

Maintenance & Community

Contribution guidelines are provided within the repository (add-a-new-agent). Specific community channels (e.g., Discord, Slack) or details on core maintainers and sponsorships are not explicitly mentioned in the README.

Licensing & Compatibility

• License: MIT-0 License.
• Compatibility: While the MIT-0 license is permissive, the project includes significant legal disclaimers. It is intended for demonstrative purposes only and is not approved for clinical use. Customers handling Protected Health Information (PHI) must ensure HIPAA compliance, potentially requiring an AWS Business Associate Addendum (BAA).

Limitations & Caveats

This toolkit is strictly for demonstrative and development purposes; it is not a substitute for professional medical advice and is not cleared for clinical use. Users are solely responsible for determining their HIPAA obligations and implementing appropriate compliance measures, including executing a BAA with AWS if necessary, before using the solution with PHI. Associated sample data and models are also not intended for production environments.