A new multi-agent AI system for market surveillance combines LangGraph for macro-level workflow orchestration and Strands for intelligent agent reasoning, deployed on AWS infrastructure using Amazon Bedrock AgentCore. The solution addresses challenges in financial services, where systems must coordinate specialized agents to analyze trading patterns, investigate suspicious activities, and generate reports while maintaining compliance and reliability. LangGraph manages state and directed graphs for multi-agent coordination, offering fine-grained control over workflow execution and shared state, with a central persistence layer supporting human-in-the-loop interactions and checkpoint-based recovery. Strands Agent serves as the reasoning engine within workflow nodes, providing model-agnostic capabilities that integrate with various large language model providers, flexible tool integration, and comprehensive observability. The release of Amazon Bedrock AgentCore last year simplified productionizing agentic solutions. The system implements state-driven workflow orchestration with LangGraph’s checkpoint system, integrates Strands agents for specialized reasoning tasks, and uses AgentCore for scalable production deployment. The complete solution is available on GitHub.