Inspiration

I came across this hackathon just one week before the submission deadline. While thinking about what to build, I started looking at my everyday online shopping experience.

I noticed how Instagram and shopping websites show advertisements and product recommendations based on things we search for and show interest in. Discovering products has become easier, but getting a better price still takes effort. We compare prices, look for discounts and sometimes negotiate ourselves.

I wondered what if an AI agent could handle the negotiation on our behalf? That thought led to NEGOZA.

What it does

NEGOZA is an AI-powered shopping agent that negotiates on behalf of buyers. Users discover a product, set their target price and maximum budget, and activate their Buyer Agent.

The Buyer Agent negotiates with a Seller Agent through offers and counteroffers to reach a mutually acceptable price. Users can then review the final deal and approve the purchase through PayPal.

How we built it

I built NEGOZA independently using React, FastAPI and Google's Gemini API. I integrated Channel3 for product discovery, KERNEL for browser automation and PayPal Sandbox for payments.

The system uses separate Buyer and Seller Agents coordinated by a negotiation engine, with Python-based validation to enforce pricing constraints.

Challenges we ran into

One of the biggest challenges was making the agents negotiate realistically. Initially, the Seller Agent sometimes accepted offers too easily or generated counteroffers above the actual listed price. I traced this to incorrect fallback merchant pricing and missing validation rules, then implemented deterministic checks in Python.

Integrating multiple external services and connecting the frontend with the backend within a limited timeframe was another challenge.

Accomplishments that we're proud of

Building a working multi-agent negotiation workflow independently, especially after discovering the hackathon so close to its deadline, was a major accomplishment.

I'm particularly proud of implementing pricing safeguards, integrating real product discovery through Channel3 and bringing together AI negotiation, browser automation and payment functionality in one application.

What we learned

This project helped me understand that building AI agents involves more than generating responses. It requires coordinating agent decisions, defining boundaries, validating outputs and handling real-world constraints.

I also gained practical experience integrating APIs, debugging backend services and building an end-to-end application under time pressure.

What's next for NEGOZA

My vision is to make NEGOZA available while people browse platforms like Instagram and shopping websites. Instead of manually searching for better prices, users could activate their shopping agent whenever they discover a product they like.

I want to explore browser extensions, wider merchant integrations and a more seamless journey from product discovery to negotiated purchase.

NEGOZA — Your AI Agent Negotiates. You Approve. It Buys.

Built With

Share this project:

Updates

Submission history