cinderline/northcinder

What it solves

NorthCinder provides an independent, user-controlled shopping agent framework that prevents marketplaces from steering buyers toward their own platforms. It ensures that product recommendations are based on neutral ranking and verified facts rather than sponsored placements or platform bias.

How it works

The system operates as a local MCP (Model Context Protocol) server that connects to an AI application. It runs a search engine locally on the user's machine, comparing products across multiple sources (such as Shopify, WooCommerce, eBay, Etsy, and Amazon) using built-in adapters. The process follows a strict research-before-recommendation workflow where the AI host creates a research plan and verifies product and seller details before ranking. To ensure security and autonomy, it requires explicit user approval for every single purchase and rejects raw credit card details, instead using opaque payment tokens or direct cart links.

Who it’s for

Users who want an AI shopping assistant that is private, self-hosted, and independent of any single marketplace's commercial interests.

Highlights

• Neutral Ranking: Sponsored offers are kept below organic results, and seller payments do not influence rankings.
• Local-First Privacy: Search history, settings, and audit logs are stored locally; the project owner does not operate a cloud service.
• Verified Research: Uses dedicated research guides for products and sellers to ensure recommendations are based on verified facts.
• Multi-Store Integration: Includes adapters for major platforms like eBay, Etsy, Shopify, and WooCommerce.
• Explicit Approval: Every checkout requires a fresh, signed approval for a specific offer and unit.