A walkthrough of building a production-ready multi-agent AI system for market surveillance using LangGraph for macro-level workflow orchestration and Strands for agent reasoning, deployed on Amazon Bedrock AgentCore. LangGraph manages state via directed graphs and checkpoint-based recovery, while Strands agents handle specialized reasoning within individual nodes. The architecture separates deterministic orchestration from localized LLM intelligence, with AgentCore providing managed runtime deployment, memory persistence via AgentCoreMemorySaver, and observability through CloudWatch and X-Ray. The pattern is applicable beyond finance to document processing, customer service, and compliance monitoring.
Table of contents
Strands: Intelligent agent reasoningLangGraph: Macro workflow orchestrationWhy LangGraph and Strands togetherAWS infrastructure and deployment with AgentCoreConclusionAbout the authors40 Impressions