Fudge MCP
Give your AI agents design taste from existing websites

What it does
Fudge MCP is a design reference engine for AI agents. It enables searching nearly 10,000 real websites by fonts, colors, components, layouts, page types, and visual similarity. It runs locally through the Model Context Protocol (MCP) and includes a Chrome extension to save references. Instead of asking AI to generate generic "modern, premium" interfaces, Fudge provides measured design evidence with screenshots.
Who it is for
This tool is primarily aimed at developers and designers who use AI coding agents (such as GitHub Copilot, Cursor, etc.) and want to improve the design output by providing real-world references. It is also useful for anyone building AI-powered design tools that need grounded design data.
Why it matters
AI agents often produce generic designs because they lack specific visual references. Fudge addresses this by giving agents access to a curated database of real website designs, searchable by visual attributes. This allows AI to generate more tailored, tasteful interfaces without relying on vague adjectives. Running locally via MCP ensures privacy and speed.
Launch signal
Fudge MCP launched on Product Hunt on the current date. It is promoted as a new tool, described as "a design reference engine for AI agents." The launch emphasizes the ability to give coding agents "better taste" by searching real websites.
Brand and naming
The name "Fudge MCP" is playful and memorable, but may confuse users unfamiliar with MCP (Model Context Protocol). The tagline clearly communicates the value proposition. The brand positions itself as a design assistant for AI agents, differentiating from generic AI design tools by grounding in real examples.
Founder
Simdi Jinkins
Related