Why did Karate replace GraalJS with its own custom JS engine (karate-js)?

Karate's JavaScript engine has changed twice in its history: originally Nashorn (bundled with the JDK), then GraalJS after Oracle deprecated Nashorn, and more recently karate-js, a JavaScript engine written from scratch specifically for Karate.

The driving reason for the second migration was concurrency. GraalJS doesn't support safely evaluating JavaScript from the same context across multiple threads at once, a limitation the Graal team has been asked about but hasn't resolved for this use case. That conflicts directly with one of Karate's central selling points: running large test suites with many scenarios executing in parallel. Every scenario touching JavaScript (which is most of them, since def, expressions, and configuration all go through the JS engine) was effectively constrained by this threading limitation.

Building karate-js from scratch let Karate Labs design specifically for thread-safe concurrent execution from the start, support modern ES6+ syntax, integrate directly with Java without a GraalVM-specific runtime dependency, and improve startup performance, all while remaining close enough to standard JavaScript that most existing feature files continue to work with minimal or no changes.