Agentic Commerce Shopping Agent is an AI shopping agent that closes the loop from a plain language request to a paid order. The user types a need such as a birthday gift under 50 dollars for a coffee lover. The AI parses the request into a structured shopping spec, searches a product catalog, and picks the best match. The agent then creates a PayPal sandbox order for that product. The user approves the payment once in the browser with a sandbox test buyer account. The agent captures the order and prints a receipt with the real PayPal transaction ID. Every step is a live call, not a mock. The sandbox order, the approval page, and the captured transaction ID are all visible in the demo video.
The agent runs as a command line flow or a Gradio web UI. Setup is a virtual environment, pip install, and three environment variables. The README gives the full run instructions so judges can reproduce the flow.
How it uses PayPal and AI
PayPal carries the money. The agent uses PayPal OAuth and the Orders v2 API in the sandbox to create orders, redirect the buyer to the PayPal approval page, and capture the approved order. The captured transaction ID on the receipt is a real PayPal order object.
AI carries the shopping judgment. GPT-6 Luna parses the free text request into a structured spec, matches products against that spec, and writes the receipt. The model is set by the OPENAI_MODEL environment variable and defaults to gpt-6-luna.
