FinSight: A Hybrid Financial-Agent System with Market Insights

The traditional financial advisory platforms have a severe limitation: the lack of real-time market information along with full user context for analysis. This paper introduces FinSight, a financial AI Agent, a production-level hybrid platform that realizes a stateful multiagent model for personalized financial intelligence. Our system is a Next.js web application that utilizes a hybrid MySQL/MongoDB persistence layer and a Python-based AI microservice that runs a LangGraph-orchestrated four-agent collaborative system. The agents Planning, Data Retrieval, Analysis, and Report Generation work on a shared state graph to produce intricate multi-step reasoning workflows. We incorporate the Yahoo Finance API via they finance library for access to free real-time market data and add context-aware personalization by way of dynamic user profile injection. Our implementation illustrates that a set of specialized agents, each with access to domain-specific tools, can collectively offer goal-aligned financial insights that a single-agent LLM architecture cannot. We describe architectural decisions, implementation details, system performance metrics, and present a complete audit that demonstrates production-ready features and areas for improvement.