I spent 18 months building an AI-native ecommerce plugin for WordPress. Here is what I learned.

A developer spent 18 months building EasyCommerce, an AI-native ecommerce plugin for WordPress that serves as a WooCommerce alternative. The plugin uses dedicated database tables for orders, products, and customers, and features a shopping agent with 23 tools for conversational commerce. The project has been stable in production since 2025 with nearly 30 releases.

Quick intro: I have been writing WordPress since 2009 and shipping products since 2017. For the last 18 months I have been building EasyCommerce , an AI-native ecommerce plugin, as a WooCommerce alternative. It has been stable and in production since 2025, nearly 30 releases in, with real stores running on it. A few decisions I would defend: 1. Dedicated database tables from day one. Orders, products and customers live in their own wp ec tables instead of wp posts and wp postmeta . More on why in a follow-up post, but the short version is that ecommerce query patterns and the post/meta model do not get along at scale. 2. AI as core, not a bolt-on. The headline feature is a shopping agent with 23 tools that can run a full conversation with a customer, search the catalogue, apply a coupon and place an order. It is agentic, not a prompt box. 3. Ship in the open . I will be posting the architecture calls, the things that broke, and the onboarding numbers here. If you build WordPress or ecommerce products: what is one architecture decision you made early that you would make again?