An autonomous Accounts Payable AI system built using Model Context Protocol, LangGraph, and FastMCP that intelligently processes invoices, makes approval decisions, and sends Slack notifications fully automated and production-ready.

Overview

This project implements a real-world AI agent system that handles end-to-end business workflows:

📥 Fetch pending invoices from database

🧠 Validate and analyze invoice data

⚖️ Auto-approve / reject / flag for review

🔔 Send Slack notifications for actions

📊 Generate real-time AP summary reports

The system combines:

MCP Server (FastMCP) → Tool execution layer

LangGraph Agent → Workflow orchestration

LLM (Groq / OpenAI) → Decision-making engine

PostgreSQL → Invoice persistence

Slack API → Notifications

Tech Stack

🧠 LangGraph

⚡ FastMCP

🤖 OpenAI / Groq LLM

🐘 PostgreSQL

🔔 Slack Webhooks

🐍 Python 3.10+

Use Cases

Enterprise Accounts Payable automation

Finance workflow orchestration

AI-driven business operations

Agent-based ERP systems

Tool-augmented LLM applications

Future Improvements

📊 Admin dashboard (React + FastAPI)

🔁 Retry + failure recovery system

📦 Batch invoice processing

🧾 Audit logs & compliance tracking

🤝 Human-in-the-loop approvals

☁️ Docker + Kubernetes deployment

👨‍💻 Author

Danish Zulfiqar

AI Engineer | LLM Systems | MCP | LangGraph | MLOps