Kore vs Orca Minecraft Datapack Generator

Kore is a Kotlin library for generating Minecraft datapacks from code. It runs during the build and produces ordinary datapack files, so the server does not need Kore or a Kore plugin. Its typed game data can catch errors in items, blocks, sounds, and commands before a datapack is reloaded. The site lists 65+ typed commands, 50+ JSON resource types, and 80+ generated registries. Kore supports commands, recipes, loot tables, advancements, custom dimensions, functions, and predicates; Kotlin functions, loops, and variables can also generate repeated content. If its typed DSL does not cover a command, function bodies can include raw command lines. Existing datapacks can be moved over gradually, with handwritten parts continuing to run alongside generated ones. Kore is available as a Maven Central dependency, and the maker recommends a JDK and Kotlin IDE. An optional Gradle plugin can build, link, and reload a datapack on save. A browser playground can generate a datapack ZIP. The project is free and open source.

Orca Minecraft Datapack Generator turns a plain-English description into a Java datapack, creating JSON and mcfunction files, pack.mcmeta, and the folder layout. It supports recipes, loot tables, advancements, gameplay functions, and world-generation tweaks. The product page says the AI validates a pack, loads it onto a test server, and can use error logs to fix functions that fail. Datapacks run on vanilla Minecraft and Paper or Spigot servers without a client-side mod loader or player installation, but they cannot add entirely new items, blocks, or mobs. The free plan includes Orca Fast, 200 starter credits, 200 credits on later UTC sign-in days, free downloads, and an 8 GB test server. Orca Play costs $6.99 a month; Orca Pro costs 9.99 USD for the first month, then $19.99 a month; Orca Max costs $99.99 a month. Orca says it scans generated builds and uploaded files for security risks. Microsoft Azure OpenAI Service processes chat transcripts and build requests; Orca says user content is not used to train models.