Why does Karate not require step definitions like traditional Cucumber?

Traditional Cucumber treats Gherkin purely as a human-readable specification layer: every step, however simple, needs a corresponding piece of glue code (a step definition) written in Java, Ruby, or whatever language the project uses, that implements what that step actually does. Karate takes a different approach: it interprets a fixed, built-in vocabulary of steps (url, path, method, match, and so on) directly, so there's nothing to implement for those common cases.

This works because Karate constrains itself to a specific domain, HTTP API testing, UI automation, and related concerns, rather than trying to be a general-purpose behavior-driven framework for any conceivable domain. Traditional Cucumber's flexibility (any step text, any glue code) is exactly what makes step definitions necessary; Karate trades some of that open-endedness for a smaller, purpose-built vocabulary that doesn't need custom implementation.

When a test genuinely needs custom logic beyond Karate's built-in steps, embedded JavaScript (via def, eval) or direct Java interop covers that gap, without requiring the ceremony of a full step-definition class for every custom need.