Inspiration
Over the years we kept running into the same frustrating experience while shopping online: buying something, only to later realize it was just a dropshipped product that could have been purchased much cheaper somewhere else. After this happened enough times, we started thinking about how much money people probably waste simply because they never see the better options that exist.
We wanted to build something that fixes that problem in the moment. Instead of finding cheaper alternatives after a purchase, we thought it would be much more useful if a tool could surface those options while you’re still on the product page deciding whether to buy.
What it does
Merx is a browser-based shopping assistant that analyzes the product page you are currently viewing and searches the web for cheaper listings or better alternatives.
Using a Gemini-powered agent, the system looks for: - The same product sold elsewhere at a lower price - Similar products with comparable specifications - Alternative items that may better suit the user’s needs
The goal is simple: help users avoid overpaying and make smarter purchase decisions. Instead of manually checking multiple sites, the assistant performs the search automatically and shows better options directly while the user is browsing.
How we built it
Merx is built as a browser extension paired with a cloud backend that handles the heavier processing and product discovery.
During development we used a combination of Claude Code and Codex to help build different parts of the extension and backend logic. The core intelligence of the system is powered by Gemini, which acts as the main agent responsible for analyzing products and searching for comparable alternatives.
Gemini handles most of the heavy lifting by: - Interpreting product data extracted from the page - Searching for related listings across marketplaces - Evaluating which products are actually comparable - Ranking the results before returning them to the extension
The extension collects product details from the page, sends them to the backend, and then displays the ranked alternatives directly to the user.
Challenges we ran into
One of the biggest challenges we faced was product ranking accuracy. Finding related products is relatively easy, but determining which results are actually good alternatives turned out to be much harder than expected.
Early on we frequently received results that technically matched but were too different from the original item to be useful. Building filtering and ranking logic that could consistently return relevant alternatives required a lot of iteration.
Another major challenge came from bot protections on many product pages. A large number of ecommerce sites block scraping attempts, which sometimes prevented us from retrieving basic data like product names or prices. Without reliable product data, recommendation accuracy drops quickly.
Data access ended up being one of the biggest technical hurdles throughout the project.
Accomplishments that we're proud of
One of the biggest accomplishments was getting the entire system working end-to-end. Early on there were a lot of moving parts — browser extension logic, APIs, Gemini integration, product discovery, and ranking — and connecting all of those pieces together was a major milestone.
The first time we successfully returned a product recommendation that passed through the entire architecture felt like a huge breakthrough.
We’re also proud of integrating marketplace APIs, especially larger ones like AliExpress. Being able to pull real product data and use it to generate meaningful comparisons was a key step toward making the assistant genuinely useful.
What we learned
This project pushed us into several areas we hadn’t worked in before, especially browser extension development. Building software that runs inside the browser while communicating with backend services and AI models introduced a lot of new challenges.
We also learned how messy product data can be. Listings across different marketplaces rarely follow consistent formats, which makes automated comparison much more complex than it initially seems.
More than anything, this project taught us how much work goes into building systems that appear simple to the end user.
What's next for Merx
There are still several areas we plan to improve as Merx continues to develop.
One issue we want to solve is price accuracy. Occasionally our results return outdated or slightly incorrect prices due to indexing delays or listing updates. Improving how we verify and refresh price data will be a big focus going forward.
Another major priority is improving our data collection capabilities. At the end of the day, the biggest limitation we face is still data access. The better the data we can gather, the better the recommendations Merx can provide.
We also plan to publish the extension to the Chrome Web Store. Right now the tool requires a developer installation, which isn’t ideal for everyday users. Making the extension easily accessible will allow anyone to benefit from smarter price comparisons while shopping online.
As we improve data access, ranking logic, and scraping reliability, we’re confident Merx will become significantly more accurate and useful over time.
Log in or sign up for Devpost to join the conversation.