Israeli development uses AI to lower your supermarket shopping cart costs
Israeli developer Nitai Aharoni has launched SuperMCP, a free tool that leverages AI to compare grocery prices across 10 major retail chains. By integrating structured data directly into AI models, the system provides accurate, real-time cost comparisons for consumers.
Source: SuperMCP
Is it possible to beat the rise in food prices with the help of language models? One Israeli developer thinks so. He has managed to bring order to the chaos of retail chains and built a tool that accurately compares your actual shopping cart. We tested it to see where it is really cheapest.
"Preventing the model from inventing a price"
Meet SuperMCP, a new free tool created by developer Nitai Aharoni that performs price comparisons for home shopping. The tool currently includes data from 10 retail chains: Osher Ad, Carrefour, Rami Levy, Shufersal, Yohananof, Tiv Taam, Keshet Teamim, Stop Market, Freshmarket, and Dabah. The system is capable of comparing prices in physical branches with online prices based on the user's location.
Unlike existing tools like Pricez, SuperMCP runs directly on your favorite AI model, including all your personal settings. If you have a subscription with Computer Use capabilities, you can even send an agent to place the order for you. It requires no user account, only a simple MCP connection.
In a conversation with Geektime, Aharoni explained that he built the tool using Claude via Cursor. He designed the architecture, chose data sources, and defined the database structure. By using MCP at the heart of the system, he prevents AI hallucinations: "Instead of the model trying to guess prices or search for them itself, it turns to SuperMCP to receive structured and up-to-date data."
The system is based on TypeScript, uses PostgreSQL and pgvector for database and search, Fastify for the server, and the official MCP SDK. While production runs in Google's cloud, the tool operates directly against the users' AI models, which reduces costs and ensures that the AI provides real facts rather than reasonable-sounding fabrications.
Scraping data in a fragmented market
"There is no single government API in Israel that centralizes all food prices in a convenient format," Aharoni explains. Each chain operates differently, ranging from XML files on FTP servers to dedicated web portals. These discrepancies in fields and encoding required building a separate adapter for each source.
"Think of it as an interpreter between many languages. Each chain speaks its own format, and SuperMCP translates them all into one language," he says. When a reliable barcode is missing, the system uses text cleaning and semantic search. It also accounts for physical versus online branches, including shipping fees and minimum order requirements.
How fresh is the price?
Aharoni emphasizes that prices are not real-time: "It is important for me to be transparent. They are based on the latest information published by the chains." The ingestion process runs once a night, and for each price, the publication and ingestion time is saved to track its freshness.
Currently, the tool is free and funded by Aharoni. While he does not rule out a commercial layer for business use in the future, the basic consumer version is not intended to become a paid service.
The tool is available via Claude, ChatGPT, Gemini CLI, Cursor, and VS Code. In our test with a 20-product cart, the model correctly identified Rami Levy as the cheapest option, though it struggled to finalize the order due to stock issues and interface limitations.
You can start with SuperMCP here, or explore the source code on GitHub.
