Bluecrop

**Tell us what you need. Bluecrop finds it, verifies it, and helps you buy it.**

Comment

Inspiration

Online shopping gives users unlimited choice but forces them to compare stores, delivery costs, product variants, and seller credibility manually. We wanted to replace this process with one guided conversation.

What we built

Sprzedawca całego internetu is an AI shopping agent that turns short user answers into a structured Shopping Brief. It then scrapes relevant offers, calculates the total cost, evaluates risk, and recommends the best purchase.

Our demo covers two scenarios:

  • an immediate purchase of an electric toothbrush;
  • a complete beginner guitar set with optional price monitoring.

How we built it

We combined a responsive frontend, an interactive questionnaire, GPT-5.6 Luna, a product database, a scraping pipeline, and backend modules for product matching, cost calculation, ranking, and risk scoring.

The model runs only after the brief is collected, reducing token usage and producing more consistent results.

Challenges and lessons

The biggest challenges were matching products across stores, comparing the real cost instead of the displayed price, and preventing the agent from exceeding the user’s budget.

We learned that a reliable shopping agent cannot be just a chatbot. It needs structured data, deterministic rules, transparent recommendations, and controlled AI autonomy.

Built With

Share this project:

Updates

Submission history