hcls-get-started

Getting Started with the HCLS Agents Toolkit

When to use this skill

• Developer asks "how do I build an HCLS agent?"
• Developer asks "what agents are available for healthcare/life sciences?"
• Developer wants to understand the toolkit's architecture and offerings

Overview

The HCLS Agents Toolkit provides three types of resources:

1. Skills — Domain workflow knowledge (this directory: skills/)
2. MCP Servers — Domain tools accessible from any MCP client (mcp-servers/)
3. Agent Catalog — 36+ production-ready deployed agents (agents_catalog/)

Choosing Your Path

Goal	Start here
Build a new HCLS agent from scratch	Use hcls-build-agent skill + agentcore_template/
Deploy an existing catalog agent	Pick from agents_catalog/, follow its README
Connect domain tools to your IDE	Configure MCP servers from mcp-servers/
Run HCLS workflows without deploying	Use skills + MCP servers directly in your assistant

Architecture

Agents are composed from:

• Skills (knowledge) — SKILL.md files that teach workflows
• MCP tools (actions) — deployed via AgentCore Gateway/Runtime or configured from public servers
• Agent reasoning (orchestration) — system prompts + Strands framework + guardrails

Key References

• Agent catalog: agents_catalog/ (36+ agents across genomics, drug discovery, clinical trials, etc.)
• Deployment templates: agentcore_template/ (backend) or FAST (full-stack)
• Framework: Strands Agents
• Infrastructure: Amazon Bedrock AgentCore

Next Steps

1. Browse skills by domain: genomics, drug discovery, clinical trials, biomarker, terminology
2. Configure MCP servers for immediate tool access (see mcp-servers/aws-public/ and mcp-servers/third-party/)
3. Use hcls-build-agent skill when ready to build a custom agent
4. Use hcls-deploy-agent skill when ready to deploy to production