FinConnectAI: Your AI Agent for Real-World Banking & SMB Insights
Author: Vikas Sahani
Demo Video: https://youtu.be/p3UMmnf9rec?si=PSe-x-Hre99i-i8y
Overview
FinConnectAI v2.1-enterprise-mcp is a modular AI agent system designed to bridge the gap between structured financial data and explainable decision-making.
This version focuses on real-world problem solving in the banking and fintech space by enabling:
Customer segmentation
Fraud and risk profiling
Conversational agents powered by LLMs
Built for individual professionals, banking teams, and AI builders looking to accelerate compliance and CRM workflows with responsible, offline-friendly AI.
Key Features
Modular Agent Architecture: Built using FastAPI, Gradio, and Hugging Face’s MCP protocol for multi-agent coordination.
Offline & Secure: Compatible with local LLMs (via Ollama) and vector databases (ChromaDB, FAISS) to keep sensitive data in-house.
Plug-and-Play Modules: Easily extend goals like “Detect Fraud” or “Summarize Sales” by editing agent JSON configs.
Developer Toolkit: Includes custom prompts, goal configuration, and explainability support.
Architecture Diagram
Gradio UI (chat/upload)
|
FastAPI
|
AgentRouter → GoalSelector
|
ModelContextProtocol (Hugging Face MCP)
|
[GPT-4 / Gemini / Ollama]
|
Insights / Logs / Explainability
Try It Now
Deploy your own version by cloning the v2.1 branch:
git clone -b 2.1 GitHub - VIKAS9793/FinConnectAI
cd FinConnectAI
pip install -r requirements.txt
python app.py
Built With
FastAPI
Gradio
Hugging Face MCP
LangChain
Ollama (optional for local LLMs)
Author Note
As a finance professional transitioning into AI, I built this project to showcase how AI agents can solve real-world compliance and customer analytics problems in SMB and banking domains.
If you’re hiring, collaborating, or building in fintech-AI, feel free to reach out or fork the repo!