Find Magic cards
by what they do
Search 25,000+ cards using TF-IDF similarity, type matching, and weighted scoring across oracle text, keywords, colors, and stats.
Learn how it works→Example searches
Find every 3-damage instant variant — Shock, Burst Lightning, Play with Fire. Get the full spectrum of direct damage options.
Discover cards with similar tax effects. The algorithm surfaces Rhystic Study, Smothering Tithe, and Mystic Remora.
Compare mana rocks by CMC, output, and card type. Understand the fast mana landscape across formats.
How scoring works
TF-IDF vectors measure semantic similarity in rules text. Cards with matching mechanics rank higher.
Jaccard distance on card types. Creature matches creature, instant matches instant.
Ability word matching. Flash, flying, trample, etc.
Identity matching plus normalized CMC, P/T comparisons.
Technical overview
Every card in Scryfall's database (~25,000 paper-legal cards) gets processed into multiple vector representations. Oracle text becomes a TF-IDF vector. Types, keywords, colors, and stats each get their own similarity functions.
When you search, your target card is scored against the entire database. The five similarity scores get weighted and combined into a final ranking. An adaptive threshold filters out weak matches.
This isn't strategic equivalence — a Counterspell won't match with Swords to Plowshares just because both are removal. It's structural similarity. Oracle text patterns, converted mana cost, card types.
Limitations
Paper-legal cards only. No Arena or MTGO exclusives in the database.
Algorithmic similarity doesn't understand context or strategy. It matches text patterns and structure, not metagame roles or strategic equivalence.
Results are a starting point for discovery and need human interpretation. The algorithm can't know what you actually need for your deck.