Hi everyone,
I’ve built an open-source MCP server for Elementary POS that allows AI assistants such as Claude, Cursor, and opencode to interact directly with the Elementary POS API.
Elementary POS already provides a REST API for integrating with external systems, including receipts, orders, items, stock, taxes and webhooks.
The goal of this project is to make that API accessible to AI agents through the Model Context Protocol (MCP).
The server currently provides 32 tools, including:
Customers — create, edit, delete and retrieve
Orders — create and retrieve
Sales items — create, edit, delete and retrieve
Stock — retrieve and update
Receipts and bills
Suppliers
Tax settings
Webhook management
It supports both local stdio connections and Streamable HTTP, so it can be used locally or deployed remotely with Docker.
🔓 Completely open source
The project is fully open source. Anyone can:
Deploy it on their own infrastructure
Use their own Elementary POS API key
Run it locally with an MCP-compatible client
Deploy it remotely with Docker
Modify or extend the available tools
GitHub repository:
https://github.com/pedrocastellanos/elementary-pos-mcp
I’m sharing it here because I think MCP could be an interesting way to build a new generation of integrations around Elementary POS — especially as AI agents become capable of performing real business operations, rather than simply answering questions.
If anyone is interested in trying it, extending it, or has ideas for additional Elementary POS API tools that should be exposed through MCP, I’d be happy to hear your feedback.
