MCP Accounting

A minimal Model Context Protocol (MCP)-style accounting analysis API built with Python and FastAPI. This project demonstrates how an AI agent or external service can interact with accounting data through structured API tools to detect financial anomalies.

The current MVP focuses on detecting unusually large transactions from accounting datasets.

Overview

This project provides a lightweight backend service that:

• Loads accounting transactions from a CSV file

• Detects unusually large transactions

• Detects potential duplicate payments

• Exposes the analysis through REST endpoints that can be used by:

  • AI agents

  • automation workflows

  • external applications

The API is designed to resemble MCP-style tool endpoints, which makes it suitable for integration with LLM-based agents.

Current Features

• Transaction ingestion from CSV

• Anomaly detection based on statistical thresholds

• Duplicate payment detection

• REST API with FastAPI

• Interactive API documentation via Swagger

• Command-line testing using curl and jq

Project Structure

mcp-accounting
│
├── app
│   ├── api
│   │   └── routes.py
│   │
│   ├── core
│   │   └── config.py
│   │
│   ├── data
│   │   └── loader.py
│   │
│   ├── mcp
│   │   └── tools.py
│   │
│   ├── models
│   │   └── schemas.py
│   │
│   ├── services
│   │   └── anomaly_detection.py
│   │
│   └── main.py
│
├── data
│   └── transactions.csv
│
├── requirements.txt
└── README.md

Installation

Clone the repository:

git clone https://github.com/<your-username>/mcp-accounting.git
cd mcp-accounting

Create a virtual environment:

python -m venv venv
source venv/bin/activate

Install dependencies:

pip install -r requirements.txt

Running the Server

Start the FastAPI server:

uvicorn app.main:app --reload

Server will run at:

http://127.0.0.1:8000

API Documentation

Swagger UI:

http://127.0.0.1:8000/docs

OpenAPI schema:

http://127.0.0.1:8000/openapi.json

Available Endpoints

Health Check

GET /health

Example:

curl http://127.0.0.1:8000/health

List Available Tools

GET /tools

Example:

curl http://127.0.0.1:8000/tools

Detect Large Transactions

POST /tools/detect_large_expenses

Example:

curl -X POST http://127.0.0.1:8000/tools/detect_large_expenses | jq

Example response:

{
  "results": [
    {
      "date": "2025-01-15",
      "vendor": "Dell",
      "amount": 8200,
      "description": "Equipment",
      "threshold": 6860.00,
      "reason": "Transaction above 95th percentile of amounts"
    }
  ]
}

Detect Duplicate Payments

POST /tools/find_duplicate_payments

Example:

curl -X POST http://127.0.0.1:8000/tools/find_duplicate_payments | jq

Sample Dataset

Example transactions.csv:

date,description,vendor,amount
2025-01-01,Office Supplies,Staples,120
2025-01-05,Consulting Fee,ABC Consulting,1500
2025-01-10,Consulting Fee,ABC Consulting,1500
2025-01-15,Equipment,Dell,8200
2025-01-20,Software License,Microsoft,300

Technology Stack

• Python

• FastAPI

• Pandas

• Uvicorn

Optional developer tools:

• curl

• jq

Development Status

Current version is an early MVP focused on core accounting anomaly detection.

Planned improvements include:

• CSV upload endpoint

• AI-generated explanations for anomalies

• Vendor spending analysis

• PostgreSQL support

• MCP-compatible tool schema definitions

• Accounting reports API

License

MIT License

Author

Edu Senior Python Developer