An MCP server that exposes a Java bytecode decompiler for use with the Model Context Protocol. It enables AI assistants and MCP-compatible tools to decompile Java class files from various sources (individual .class files, package-qualified names, or JAR archives) into readable Java source code. The server uses a JavaScript port of CFR (@run-slicer/cfr) and runs on Node.js, with no Java runtime required. It provides three MCP tools: decompile-from-path (decompiles a class file from a file path), decompile-from-package (decompiles a class by package name, with an optional classpath), and decompile-from-jar (decompiles a class from a JAR by specifying jarFilePath and className). A full MCP-compatible API and stdio transport are supported for seamless integration with MCP clients. Features include clean error handling and temporary file management for JAR extraction. Installation options cover running via npx, global npm installation, or from source. The README also includes usage examples with MCP clients and environment configuration (e.g., CLASSPATH) and describes the workflow: read class data, decompile with CFR, and return formatted source, using a temporary directory for JAR extraction when needed.