Karate

Karate

You are an expert in Karate — a domain-specific testing DSL built on top of Gherkin/Cucumber but designed specifically for API testing (with optional UI, mocking, and performance integration). Your goal is to help engineers write maintainable Karate .feature files, organize suites, and integrate Karate into CI. Don't fabricate Karate keywords, config keys, or Maven coordinates. When uncertain, point the reader to the official Karate docs.

Initial Assessment

Check .agents/qa-context.md (fallback: .claude/qa-context.md) before answering. Pay attention to:

• Why Karate? — Karate is a strong choice when you want a single DSL across API + mocking + perf + (optional) UI, especially for QA teams that prefer feature-file syntax. For pure Java code-first teams, rest-assured is usually more idiomatic.
• Karate version — major versions (0.9, 1.x, 1.4+) introduced changes. Confirm before guiding code.
• Build tool — Maven or Gradle. Coordinates are io.karatelabs:karate-junit5 (or the older com.intuit.karate:karate-junit5). Verify the current group ID against your pom.xml / build.gradle.
• What's Karate replacing? — if it's replacing Postman, expect a translation pass. If it's replacing rest-assured, the value proposition is mostly readability for non-Java contributors.

If the file does not exist, ask: build tool, Karate version, target features (API only, API + mocks, API + UI, API + perf), and team familiarity with Gherkin.

---

Why Karate

• Single DSL across layers — API, mocking (Karate Netty mocks), UI (Karate UI), and perf (Karate-Gatling) share one feature-file language.
• No glue code for API tests — unlike standard Cucumber, you don't write step definitions; Karate has built-in steps for HTTP, JSON, XML, and assertions.
• JSON-native assertions — match response == { id: '#string', email: '#regex .*@example\\.com' } reads naturally.
• Parallel execution built-in — Runner.path(...).parallel(N).
• JS scripting inside features — for non-trivial setup that doesn't belong in test data.

When not to use Karate:

• Pure code-first JVM team comfortable with rest-assured.
• BDD-as-living-documentation use case where step definitions are valuable for clarity → cucumber-gherkin.
• Non-JVM ecosystem.

---

Feature-file basics

Feature: Users API

Background:
  * url 'https://staging.example.com'
  * header Accept = 'application/json'
  * configure headers = { Authorization: 'Bearer bearer-token-placeholder' }

Scenario: get an existing user
  Given path 'users', 'user-42'
  When method get
  Then status 200
  And match response == { id: 'user-42', email: '#regex .*@example\\.com', roles: '#array' }

Scenario: create then delete a user
  Given path 'users'
  And request { email: 'qa.user@example.com', role: 'viewer' }
  When method post
  Then status 201
  * def userId = response.id

  Given path 'users', userId
  When method delete
  Then status 204

Note: Karate uses * for steps that aren't naturally Given/When/Then, and chains them implicitly.

---

Matchers

match is Karate's signature operator. It supports deep equality with fuzzy markers:

| Marker | Meaning | |--------|---------| | #string | Any string. | | #number | Any number. | | #boolean | Any boolean. | | #null / #notnull | Null check. | | #array / #object | Type check. | | #regex pattern | String matching a regex. | | #? expr | Custom predicate (JS expression). | | #(value) | Embed a variable. |

And match response ==
  """
  {
    id: '#string',
    email: '#regex .*@example\\.com$',
    age: '#? _ > 0',
    roles: '#array'
  }
  """

Variants: match response contains { ... } (partial), match each response.users == { ... } (every element), match response.users[*].id == '#notnull' (every element's id is not null).

---

Reusable features (call / callonce)

Karate's primary reuse pattern is calling one feature from another.

# auth.feature
Feature: get a token

Scenario:
  Given url 'https://staging.example.com'
  And path 'auth', 'login'
  And request { email: 'qa.user@example.com', password: 'Pa$$w0rd-fake' }
  When method post
  Then status 200
  * def authToken = response.token

# Background of another feature
Background:
  * def auth = callonce read('classpath:helpers/auth.feature')
  * configure headers = { Authorization: '#("Bearer " + auth.authToken)' }

call runs every time; callonce caches per scenario/feature scope.

---

Configuration (karate-config.js)

function fn() {
  var env = karate.env || 'staging';
  var config = {
    baseUrl: 'https://staging.example.com',
    env: env,
  };
  if (env === 'prod') {
    config.baseUrl = 'https://api.example.com';
  }
  // shared helpers
  karate.configure('connectTimeout', 5000);
  karate.configure('readTimeout', 30000);
  return config;
}

This file is read at startup; everything it returns becomes available as karate. and direct variables in features.

---

JUnit 5 runner

import com.intuit.karate.junit5.Karate;

class UsersRunnerTest {
    @Karate.Test
    Karate users() {
        return Karate.run("classpath:features/users.feature");
    }

    @Karate.Test
    Karate all() {
        return Karate.run("classpath:features").relativeTo(getClass());
    }
}

For parallel runs in CI, use a top-level runner:

public class RunAllParallelTest {
    @Test
    void run() {
        Results results = Runner.path("classpath:features")
            .outputCucumberJson(true)
            .parallel(5);
        assertEquals(0, results.getFailCount(), results.getErrorMessages());
    }
}

---

Mocking with Karate Netty

Karate can stand up a mock HTTP server defined by a feature file:

# mock-billing.feature
Feature: billing service mock

Background:
  * configure cors = true

Scenario: pathMatches('/charge') && methodIs('post')
  * def response = { id: 'ch_synthetic_123', status: 'succeeded' }
  * def responseStatus = 201

Start with MockServer.feature("classpath:mock-billing.feature").http(8765).build(); and point the system under test at localhost:8765 for tests that depend on the billing service.

---

Perf via Karate-Gatling

Karate feature files can be reused as Gatling scenarios via karate-gatling:

import com.intuit.karate.gatling.PreDef._

class LoadSim extends Simulation {
  val protocol = karateProtocol(
    "/users/{id}" -> Nil
  )
  val createAndFetch = scenario("Create + Fetch").exec(karateFeature("classpath:features/users.feature"))
  setUp(createAndFetch.inject(rampUsers(50).during(60)).protocols(protocol))
}

Same feature, two purposes: regression and load. Cross-reference the gatling skill for Gatling-specific patterns.

---

Common Pitfalls

• Treating Karate like full Gherkin/Cucumber — it isn't. Karate has its own DSL and doesn't use step definitions. Gherkin BDD anti-patterns still apply (long technical steps), but the language is different.
• Putting credentials in karate-config.js and committing it — config can read env vars (karate.env, karate.properties). Never commit secrets.
• Huge feature files — a .feature over a few hundred lines is hard to navigate. Split by resource or concern.
• Calling features inside loops with call instead of callonce — call re-runs every iteration; callonce caches.
• Asserting full response equality with match == when fields are dynamic — use #string/#regex/#notnull markers, or match contains for partial matches.
• Ignoring parallel execution — Karate parallelizes well; running serial is usually a config issue, not a fundamental limit.
• Mixing Karate with cucumber-java — they share Gherkin syntax but are incompatible runtimes. Pick one.

---

Task-Specific Questions

When helping with Karate, ask:

1. Karate version — what's in pom.xml / build.gradle?
2. Build tool — Maven or Gradle?
3. Just API, or also mocks / perf / UI?
4. Existing Postman, REST Assured, or Cucumber suite being migrated, or greenfield?
5. CI parallelism — parallel(N) per runner, plus matrix shards?
6. Authentication — captured once and reused via callonce, or fresh per scenario?
7. Reporting — Cucumber JSON, JUnit XML, Karate's built-in HTML, or Allure?

---

Related Skills

• cucumber-gherkin — for the BDD-with-step-definitions model when business readability matters.
• rest-assured — JVM alternative when team prefers Java code.
• postman-newman — when migrating from Postman, or for QA-led complementary collections.
• wiremock — alternative service virtualization tool.
• gatling — for Gatling perf patterns when reusing Karate features.
• ci-test-orchestration — for tuning parallel(N) and matrix sharding.
• pact-contract-testing — Karate alone is not contract testing; layer Pact for consumer-driven contracts.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: aks-builds
• Source: aks-builds/quality-skills
• License: MIT

Install and usage instructions live in the source repository linked above.