FastAPI + MCP financial agent with autonomous orchestration and client-side tool calling with Next.js frontend, built for the Google Hackathon.
Developed a financial AI agent platform combining FastAPI backend with MCP (Model Context Protocol) for tool calling and a Next.js frontend for client-side orchestration. The system features autonomous agent planning, dynamic tool selection, and real-time financial analysis, built for the Google AI Hackathon.
Key Features
Autonomous agent orchestration with MCP tool calling
Client-side tool execution via Next.js frontend
Real-time stock analysis and portfolio insights
Streaming chat with dynamic tool-calling
FastAPI backend with PostgreSQL persistence
GCP deployment for production readiness
Technology Stack
Frontend
Next.jsReactTypeScriptTailwindCSS
Backend
FastAPIPython
Database
PostgreSQL
AI/ML
MCPAgentsTool CallingGemini
Tools
GCPDockerGit
Challenges
Designing MCP-based tool calling across client and server
Autonomous orchestration with reliable tool selection
Streaming responses with parallel tool execution
Key Learnings
MCP protocol and client-side tool calling patterns