Team 46. eBaySnipe

Inspiration

We are college students trying to move apartments on a tight budget. We kept missing amazing deals on furniture and electronics because they were snapped up before we even saw them. We needed something smarter than just setting alerts — we wanted an AI agent that could watch the listings for us and act the moment something good appeared. That is what inspired us to build eBaySnipe.

What it does

eBaySnipe is an autonomous AI agent that monitors eBay in real time. You give it a voice or text prompt like "vintage Sony camcorder under 200" or "rare Magic the Gathering cards" and it instantly starts scanning new listings. It understands your intent, evaluates listings for value and rarity, and surfaces matches as soon as they appear. No refreshing. No filters. No waiting. The agent watches, thinks, and acts for you.

How we built it

We used openai and minimax to power the voice agent, turning flexible user intent into structured search goals. We used Browserbase with Anthropic to browse and extract listing that matches the user requirements. All components were tied together in an agent loop that listens for new deals, reasons over their value, and immediately surfaces the best ones to the user.

Challenges we ran into

At first we wanted to built this for facebook marketplace, but we ran into many problems during login. The agent wasn't smart enough to navigate around fb's multistep verification, so we had to pivot to ebay deals. Tuning the browser base agent to understand varied user prompts and extract precise constraints required careful prompt engineering. Integrating live scraping with agent decision making and keeping response times low was difficult under a tight hackathon timeline.

Accomplishments that we're proud of

We built a fully autonomous agent that actually works. It is able to understand flexible user input, search the real world in real time, and return genuinely valuable results. We also successfully integrated multiple APIs including voice processing, web scraping, and autonomous agent control all within just a few hours.

What we learned

Building agentic flows is not just about connecting language models. It is about chaining the right actions together and giving the AI the ability to reason, filter, and act like a human would. We also learned how to connect real-world search with voice commands and the value of using tools like Apify and Browserbase for real-time decision making.

What's next for eBaySnipe

We want to set up 30 minute or custom interval agents that can keep checking for new listings while you sleep. Imagine waking up to a curated list of perfect matches waiting in your inbox all found by your personal AI.

Built With

• anthropic
• browserbase
• claude
• css
• javascript
• minimax
• tailwind
• typescript