ERP & File MCP Server

A dual FastMCP (Model Context Protocol) server built with FastAPI. This application exposes two concurrent MCP servers natively routed via custom ASGI middleware to interact with backend ERP APIs and local/Google Sheets data.

Features

• Dual MCP Endpoints:

  • /mcp: Connects to an external ERP API, caches data locally into PostgreSQL with adaptive fetching algorithms, and exposes tools for querying standard business reports.

  • /mcp-file: Loads data from a Google Sheet (if GOOGLE_CREDENTIALS provided) or a local file (Excel/CSV), exposing tools to list, search, and summarize data dynamically.

• Adaptive Database Sync: Heavily optimizes API usage by scaling fetch windows based on historical data density. Keeps a rolling cache up-to-date automatically using background concurrency.

• Robust Error Handling: Handles network, API, and credential failures seamlessly, preventing server crashes and falling back when sources become unreachable.

Related MCP server: OpenXE MCP Server

Endpoints / Test Paths

• ERP Data FastMCP: http://localhost:8080/mcp

• File / Sheets FastMCP: http://localhost:8080/mcp-file

Prerequisites

• Python 3.11+

• PostgreSQL

• Valid .env configuration (see .env.example)

Environment Variables

Check .env.example for full configuration details. Key variables include:

• DATABASE_URL: PostgreSQL connection string.

• GOOGLE_CREDENTIALS: JSON string representation of a Google Service Account (optional, for Google Sheets integration).

Running Locally

1. Install dependencies:

pip install -r requirements.txt

2. Start the Uvicorn server:

uvicorn main:app --host 0.0.0.0 --port 8080

Docker

Build and run using Docker:

docker build -t erp-mcp-server .
docker run -p 8080:8080 --env-file .env erp-mcp-server