andridejager.me
All projects

February 2025

WebMCP

A sample ecommerce site that can be controlled entirely through an AI assistant. Browse, filter, add to cart, and checkout via natural language rather than the UI.

MCP AI Agents Ecommerce JavaScript LLMs

WebMCP is a proof-of-concept that demonstrates what ecommerce looks like when the interface is natural language rather than a graphical UI.

Built on a sample store, the AI assistant has full control of the shopping experience. Users can browse by category, filter by price or attribute, add items to their cart, and complete checkout entirely through conversation. No clicks required.

The project was built to explore the MCP specification in a browser context ahead of anticipated native support, and to validate how well current LLMs can handle the state management required for a multi-step transactional flow. The results were compelling enough to warrant further investment in the approach.