ComputeFinnie — AI Finance Assistant
A production multi-agent AI system with six specialist agents for financial education, portfolio analysis, live market data, goal planning, news synthesis, and tax guidance.
Overview
Finnie is a conversational AI finance assistant built on a six-agent LangGraph system. Instead of a single general-purpose model trying to answer everything, a router dispatches each query to the specialist agent with the right tools and context.
The six agents
- Finance Q&A — educational concepts grounded via RAG over a 50+ article knowledge base
- Portfolio Analysis — holdings breakdown, diversification scoring, live price enrichment
- Market Analysis — real-time stock and index data via yFinance + Alpha Vantage
- Goal Planning — retirement/savings projections across conservative, moderate, and aggressive scenarios
- News Synthesizer — thematic aggregation and summary of financial headlines
- Tax Education — account types, 2024 contribution limits, IRS rules
Architecture
Built on LangGraph's StateGraph with shared conversation state and intent-based routing. The Gemini 2.0 Flash backbone keeps inference fast and cost-efficient across many short agent calls per conversation.
Claude Desktop integration
The system doubles as an MCP server exposing seven tools — stock quotes, portfolio analysis, goal calculations, concept explanations, and more — directly inside Claude Desktop.