10ML Perfume Finder

10ML Perfume Finder is a product comparison tool. A free, keyless MCP server that plugs fragrance data into Claude, ChatGPT, Cursor, VS Code and other assistants. It exposes 15 read-only tools for searching 6,000+ perfumes, comparing scents, finding similar bottles and building recommendations, and answers in ten languages.

10ML Perfume Finder is a free MCP server that gives AI assistants a working knowledge of fragrance. Once the single endpoint is added to a client such as Claude, ChatGPT, Cursor, VS Code or Gemini CLI, the assistant can call 15 read-only tools while it answers. The underlying Fragrance Index is described as holding more than 6,000 perfumes, each with its house, year, family, accords, note pyramid, seasons and occasions. No API key, account or sign-up is required, and the tools set no cookies and keep no request history. The tools map to the questions people actually ask about scent. search_perfumes finds bottles by name, house, note, accord, family, season or occasion using plain language. get_perfume returns a full profile, compare_perfumes puts two to four fragrances side by side with shared notes and differences, and find_similar_perfumes and find_perfume_alternatives surface scents that resemble a favourite, with the reasoning behind each match. recommend_perfumes builds picks from stated likes and dislikes plus season and occasion, while pick_for_occasion and find_gift handle narrower briefs. A second group is educational: explain_note, explain_perfume_term, get_house, list_houses and suggest_layering cover vocabulary, houses and layering combinations. search_guides pulls from 10ML written guides and accepts a locale across ten languages, and non-English note names are understood too. It suits anyone who already asks an assistant for scent advice and wants sourced answers instead of guesses: shoppers comparing designer releases, collectors hunting niche houses, and gift buyers who do not know where to start. Setup is a URL paste in most clients, with one click installs for Cursor and VS Code and a stdio package for Windsurf, Cline, Goose and Zed. 10ML also publishes its own 40-question benchmark claiming the highest accuracy and lowest latency among keyless perfume MCP servers, though that comparison is self-reported. A REST API is offered separately for code that needs the raw data. Where it fits: it is a lookup and reasoning layer, not a shop. Larger indexes exist for rare and obscure bottles, so breadth is the trade-off against speed and accuracy. Lists are paginated, fair-use rate limiting may apply, and heavy jobs should be cached, since REST responses carry a five-minute cache header.