Inspiration
Running a small online shop means juggling customer issues, inventory, pricing, and payments—often alone. We built ShopCrew to give solo sellers a team of AI agents for the daily work while keeping the owner in control of important decisions.
How we built it
ShopCrew simulates a candle shop where agents forecast inventory, suggest prices, and handle refund requests. A policy engine checks each proposed action against spending limits and safety rules. Low-risk actions can proceed automatically, while others wait for approval in a live AG Grid control room. PayPal handles checkout, refunds, and supplier payouts in the sandbox, with verified webhooks updating their status.
Challenges and what we learned
The hardest part was making agent actions useful without letting them move money unchecked. We separated each agent’s proposal from the policy decision and PayPal execution, giving every action a clear reason and audit trail. We also learned that statistical models are a better fit than LLMs for forecasting inventory and estimating price elasticity.
Sales history, competitor prices, and customer complaints are simulated. Payment flows use PayPal sandbox credentials when configured.
Built With
- ag
- fastify
- node.js
- paypal
- react
- typescript
