Client java

DISCLAIMER: We use Google Analytics for sending anonymous usage information such as agent's and client's names, and their versions after a successful launch start. This information might help us to improve both ReportPortal backend and client sides. It is used by the ReportPortal team only and is not supposed for sharing with 3rd parties.

The latest version: 5.4.14. Please use Maven Central link above to get the client.

JVM-based clients configuration

How to provide parameters

There are several ways to load parameters. Be aware that higher sources override lower ones. For example, properties from file can be overridden by JVM variables.

Order	Source
1	JVM arguments
2	Environment variables
3	Properties file

JVM arguments

ReportPortal client does not necessarily need properties file to be configured. One of the option is to use JVM arguments which have the highest priority among configuration ways. To use them you need to specify them in command line after Java executable using -D flag. Example:

$ java -Drp.endpoint=https://rp.epam.com/ -jar my-tests.jar

Environment variables

In case of bypassing parameters through environment variables they should be specified in UPPERCASE separated by underscores (_). E.G.:

• rp.endpoint --> RP_ENDPOINT
• rp.skipped.issue --> RP_SKIPPED_ISSUE

Property file

The most common way to start using an agent is to copy your configuration from UI of ReportPortal at User Profile section or configure property file reportportal.properties in the following format:

rp.endpoint=https://rp.epam.com/
rp.api.key=8967de3b-fec7-47bb-9dbc-2aa4ceab8b1e
rp.launch=launch-name
rp.project=project-name
## OPTIONAL PARAMETERS
rp.reporting.async=true
rp.reporting.callback=true
rp.enable=true
rp.description=My awesome launch
rp.attributes=key:value;value
rp.rerun=true
rp.rerun.of=ae586912-841c-48de-9391-50720c35dd5a
rp.convertimage=true
rp.mode=DEFAULT
rp.skipped.issue=true
rp.batch.size.logs=20
rp.keystore.resource=<PATH_TO_YOUR_KEYSTORE>
rp.keystore.password=<PASSWORD_OF_YOUR_KEYSTORE>

For detailed parameter description see below sections.

Default properties file should have reportportal.properties name. It can be situated in the class path (in the project directory) and if client can’t find the file it logs a warning. But you can also use your custom property file specifying file's path in rp.properties.path system property or RP_PROPERTIES_PATH environment variable. The first option has priority, so if you specify the path in both system properties and environment variables then system property value will be used.

Parameters

Common parameters

Property name	Type	Description	Required
rp.endpoint	String	URL of web service, where requests should be send	Yes
rp.api.key	String	Api token of user. Required if OAuth 2.0 authentication is not used.	Conditional
rp.oauth.token.uri	String	OAuth 2.0 token endpoint URL for password grant authentication. Required if API key is not used.	Conditional
rp.oauth.username	String	OAuth 2.0 username for password grant authentication. Required if OAuth 2.0 is used.	Conditional
rp.oauth.password	String	OAuth 2.0 password for password grant authentication. Required if OAuth 2.0 is used.	Conditional
rp.oauth.client.id	String	OAuth 2.0 client identifier. Required if OAuth 2.0 is used.	Conditional
rp.oauth.client.secret	String	OAuth 2.0 client secret. Optional for OAuth 2.0 authentication.	No
rp.oauth.scope	String	OAuth 2.0 access token scope. Optional for OAuth 2.0 authentication.	No
rp.oauth.use.proxy	Boolean	Default: true Determines if OAuth HTTP client Inherits proxy settings from general client	No
rp.launch	String	A unique name of Launch (Run). Based on that name a history of runs will be created for particular name	Yes
rp.project	String	Project name to identify scope	Yes
rp.launch.uuid	String	A unique Launch UUID to which the whole test execution will be uploaded.	No
rp.launch.uuid.creation.skip	Boolean	Do not create new launch and report to predefined Launch provided by UUID above. Default true.	No
rp.launch.uuid.print	Boolean	Enables printing Launch UUID on test run start. Default false.	No
rp.launch.uuid.print.output	Enum	Launch UUID print output. Default stdout. Possible values: stderr, stdout.	No
rp.enable	Boolean	Enable/Disable logging to ReportPortal: rp.enable=true - enable log to RP server. Any other value means 'false': rp.enable=false - disable log to RP server. If parameter is absent in properties file then automation project results will be posted on RP.	No
rp.description	String	Launch description	No
rp.attributes	String	Set of attributes for specifying additional meta information for current launch. Format: key:value;value;build:12345-6. Attributes should be separated by “;”, keys and values - “:”.	No
rp.reporting.async	Boolean	Enables asynchronous reporting. Available values - true (by default) or false. Supported only in 5+ version.	No
rp.reporting.callback	Boolean	Enables callback reporting. Available values - true or false(by default). Supported only in 5+ vesion	No
rp.rerun	Boolean	Enables rerun mode. Available values - true or false(by default). Supported only in 5+ version	No
rp.rerun.of	String	Specifies UUID of launch that has to be rerun.	No
rp.convertimage	Boolean	Colored log images can be converted to grayscale for reducing image size. Values: ‘true’ – will be converted. Any other value means ‘false’.	No
rp.mode	Enum	ReportPortal provides possibility to specify visibility of executing launch. Currently two modes are supported: DEFAULT - all users from project can see this launch; DEBUG - all users except of Customer role can see this launch (in debug sub tab). Note: for all java based clients (TestNG, Junit) mode will be set automatically to "DEFAULT" if it is not specified.	No
rp.skipped.issue	Boolean	ReportPortal provides feature to mark skipped tests as not 'To Investigate' items on WS side. Parameter could be equal boolean values: true - skipped tests considered as issues and will be marked as 'To Investigate' on ReportPortal. false - skipped tests will not be marked as 'To Investigate' on application.	No
rp.batch.size.logs	Integer	Put logs into batches of specified size in order to rise up performance and reduce number of requests to server. Default = 10	No
rp.batch.payload.limit	Long	Limit batches by payload size to avoid request rejection due to server limitations.	No
rp.rx.buffer.size	Integer	Internal queue size for log processing, increase this value along with log batch size if you see not all your logs passing to server. Default = 128	No
rp.keystore.resource	String	Keystore file path to be used in HTTPS communication	No
rp.keystore.password	String	Access password for certificate storage package, mentioned above	No
rp.keystore.type	String	Keystore type. Default: JKS	No
rp.truststore.resource	String	Truststore file path to be used in HTTPS communication	No
rp.truststore.password	String	Access password for certificate storage package, mentioned above	No
rp.truststore.type	String	Truststore type. Default: JKS	No

Launch name sets once before first execution, because in common launch parts are fixed for a long time. By keeping the same launch name we will know a fixed list of suites behind it. That will allow us to have a history trend. On Report Portal UI different launch iterations will be saved with postfix "#number", like "Test Launch #1", "Test Launch #2" etc.

Authentication

ReportPortal supports two authentication methods:

1. API Key authentication (default) - using rp.api.key parameter
2. OAuth 2.0 Password Grant authentication - using OAuth parameters (rp.oauth.*)

Authentication priority:

• If both API key and OAuth parameters are provided, OAuth 2.0 authentication will be used.
• Either API key or complete OAuth 2.0 configuration is required to connect to ReportPortal.

OAuth 2.0 configuration example:

rp.endpoint=https://reportportal.example.com/
rp.oauth.token.uri=https://reportportal.example.com/uat/sso/oauth/token
rp.oauth.username=my-username
rp.oauth.password=my-password
rp.oauth.client.id=client-id
rp.oauth.client.secret=client-id-secret
rp.oauth.scope=offline_access
rp.launch=launch-name
rp.project=project-name

Note: rp.oauth.client.secret and rp.oauth.scope are optional parameters.

If mandatory parameters are missed client will throw an InternalReportPortalClientException.

Multi-process join parameters

Property name	Type	Description
rp.client.join	Boolean	Default: true Enable / Disable multi-process launch join mode
rp.client.join.mode	Enum	[FILE, SOCKET], Default: FILE Which mechanism will be used to join multi-process launches: FILE - the client will create a locking file SOCKET - the client will open a socket
rp.client.join.port	Integer	Default: 25464 If client join mode set to SOCKET, this property controls port number of the socket
rp.client.join.timeout.value	Integer	Default: 1.8M milliseconds (30 minutes) Timeout value for secondary launches. Primary launch will wait that amount of time after test execution for secondary launch finish.
rp.client.join.timeout.unit	Enum	Default: MILLISECONDS Timeout value time unit. Should be one of values from java.util.concurrent.TimeUnit class
rp.client.join.file.lock.name	String	Default: reportportal.lock A name of a main lock file, can be an absolute path. A client which managed to obtain that lock count itself as a primary launch process. It rewrites synchronization file with its launch ID.
rp.client.join.file.sync.name	String	Default: reportportal.sync A name of a launch ID synchronization file, can be an absolute path. Each client waits for a lock on that file to get a launch ID (first line) and write its own ID to the end of the file.
rp.client.join.lock.timeout.value	Integer	Default: 1 minute Files lock / connection timeout for launches.
rp.client.join.lock.timeout.unit	Enum	Default: MILLISECONDS Timeout value time unit. Should be one of values from java.util.concurrent.TimeUnit class

HTTP parameters

Property name	Type	Description
rp.http.proxy	String	A URL of a HTTP proxy to connect to the endpoint.
rp.http.proxy.username	String	A username for used proxy, works only if Proxy URL is set.
rp.http.proxy.password	String	Password for proxy, works only if Proxy URL and Proxy Username are set.
rp.http.logging	Boolean	Default: false Enable / Disable HTTP logging.
rp.http.timeout.call.value	Integer	Default: Infinitive Timeout value for the entire call: resolving DNS, connecting, writing the request body, server processing, and reading the response body. If the call requires redirects or retries all must complete within one timeout period.
rp.http.timeout.call.unit	Enum	Default: MILLISECONDS Timeout value time unit. Should be one of values from java.util.concurrent.TimeUnit class
rp.http.timeout.connect.value	Integer	Default: 10 seconds Connect timeout for new HTTP connections.
rp.http.timeout.connect.unit	Enum	Default: MILLISECONDS Timeout value time unit. Should be one of values from java.util.concurrent.TimeUnit class
rp.http.timeout.read.value	Integer	Default: 10 seconds Data read timeout for new HTTP connections.
rp.http.timeout.read.unit	Enum	Default: MILLISECONDS Timeout value time unit. Should be one of values from java.util.concurrent.TimeUnit class
rp.http.timeout.write.value	Integer	Default: 10 seconds Data write timeout for new HTTP connections.
rp.http.timeout.write.unit	Enum	Default: MILLISECONDS Timeout value time unit. Should be one of values from java.util.concurrent.TimeUnit class

Truncation and sanitization parameters

Property name	Type	Description
rp.truncation.field	Boolean	Default: true Toggle certain field truncation to avoid API failures.
rp.truncation.replacement	String	Default: ... Replacement pattern for truncated fields
rp.truncation.item.name.limit	Integer	Default: 1024 Maximum item names length before truncation.
rp.truncation.attribute.limit	Integer	Default: 128 Maximum attribute key and value limit (counts separately)
rp.truncation.exception	Boolean	Default: true Toggle Stack Trace truncation of exceptions that being logged to ReportPortal.
rp.attribute.limit	Integer	Default: 256 Maximum number of attributes sent in request.
rp.sanitization.replace.binary	Boolean	Default: true Toggle replacement of basic binary characters with \uFFFD char.

Bug Tracking System parameters

Property name	Type	Description
rp.bts.project	String	Bug Tracking System Project name to use along with @ExternalIssue annotation. Should be the same as in corresponding integration.
rp.bts.url	String	Bug Tracking System base URL. Should be the same as in corresponding integration.
rp.bts.issue.url	String	Bug Tracking System URL Pattern for Issues. Use {issue_id} and {bts_project} placeholders to mark a place where to put Issue ID and Bug Tracking System Project name.
rp.bts.issue.fail	Boolean	Default: true Fail tests marked with @Issue annotation if they passed. Designed to not miss the moment when the issue got fixed but test is still marked by annotation.

Proxy configuration

ReportPortal supports 2 options for setting Proxy configuration:

• JVM arguments (-Dhttps.proxyHost=localhost)
• reportportal.properties file

JVM arguments

ReportPortal uses OkHttp as HTTP client, which can pick up JVM proxy settings. This is the most flexible and preferable way to configure proxies, since it supports different proxy types. You can find out more about JVM proxies on Java networking and proxies page.

Properties file

If you need to set up just a simple HTTP proxy you can use reportportal.properties file. rp.http.proxy parameter accepts HTTP proxy URL. This parameter can override JVM proxy arguments, so watch your back.

Client usage

When to use the client

This library is not an end-to-end integration with a test framework. If you use a common framework (JUnit 4/5, TestNG, Cucumber, etc.) prefer one of the ready-made agents which handle the full lifecycle for you. You can find the full list of supported Java agents by searching for the agent-java- prefix on our GitHub organization page: https://github.com/reportportal?q=agent-java-&type=all.

Use the Client directly only if:

• you need to report data from a custom or a rare test framework which is not covered by an existing agent;
• you are extending an existing agent and need to send extra items, logs or attachments;
• you are writing an auxiliary listener (for example, a browser / WebDriver / HTTP step logger) which runs together with an already running agent.

There are two typical ways to work with the client:

1. Simplified access — reuse a Launch instance that is already created and managed by an agent. This is the way to go for listeners, loggers and custom steps running inside a test which is already reported by an agent.
2. Direct instantiation — build a ReportPortal object, start a Launch from scratch, and drive the whole reporting lifecycle yourself. This is what you need when you integrate a custom or exotic framework.

Simplified usage: reuse the Launch created by an agent

When a ReportPortal agent is active, it stores the current Launch instance in a thread-local-like holder. You can retrieve it from anywhere in your code via the static method Launch.currentLaunch():

import com.epam.reportportal.service.Launch;

Launch launch = Launch.currentLaunch();
if (launch == null || launch == Launch.NOOP_LAUNCH) {
    return; // no active ReportPortal agent, nothing to do
}

The NOOP_LAUNCH check is important: when ReportPortal is disabled (for example, via rp.enable=false or because of a misconfiguration), the agent installs a no-op instance instead of throwing an exception. Your listener must gracefully skip reporting in such case.

Once you have a Launch, you can:

• inspect runtime configuration via launch.getParameters() (returns ListenerParameters);
• access the low-level REST client via launch.getClient() (returns ReportPortalClient);
• obtain the current Launch UUID via launch.getLaunch() (returns Maybe<String>);
• reuse the current-item stack via launch.getStepReporter() (returns StepReporter);
• start / finish nested items on the current item via launch.startTestItem(...) / launch.finishTestItem(...);
• send logs via ReportPortal.emitLog(...).

Example 1. Reporting the current Launch URL

Useful in a CI log, typically called from an @AfterClass / @AfterSuite hook:

import com.epam.reportportal.listeners.ListenerParameters;
import com.epam.reportportal.service.Launch;

import static java.util.Optional.ofNullable;

Launch launch = ofNullable(Launch.currentLaunch())
        .filter(l -> l != Launch.NOOP_LAUNCH)
        .orElseThrow(() -> new IllegalStateException("Launch not found"));

ListenerParameters parameters = launch.getParameters();
String launchUuid = launch.getLaunch().blockingGet();
String baseUrl = parameters.getBaseUrl();
baseUrl = baseUrl.endsWith("/") ? baseUrl.substring(0, baseUrl.length() - 1) : baseUrl;
LOGGER.info("Launch URL: {}/ui/#{}/launches/all/{}", baseUrl, parameters.getProjectName(), launchUuid);

Example 2. Resolving numeric Launch ID via the REST client

launch.getClient() exposes the same Retrofit-based REST client that the agent uses internally, so you can call any ReportPortal API method without creating another HTTP client:

ofNullable(Launch.currentLaunch()).ifPresent(l -> {
    String launchUuid = l.getLaunch().blockingGet();
    LaunchResource info = l.getClient().getLaunchByUuid(launchUuid).blockingGet();
    LOGGER.info("Launch ID: {}", info != null ? info.getLaunchId() : null);
});

Note: REST calls may throw retrofit2.HttpException if the launch is not yet visible on the server side in async mode. Wrap such calls in a small retry loop if you depend on their result.

Example 3. Reporting custom steps and attachments from a listener

The StepReporter obtained via launch.getStepReporter() maintains its own stack of "virtual" steps which are nested inside the current test item of the agent. It is the recommended entry point for any custom step / log / screenshot reporting — for instance, a Selenide, Selenium, REST Assured or HTTP client listener:

import com.epam.reportportal.listeners.ItemStatus;
import com.epam.reportportal.listeners.LogLevel;
import com.epam.reportportal.message.ReportPortalMessage;
import com.epam.reportportal.service.Launch;
import com.epam.reportportal.service.ReportPortal;
import com.epam.reportportal.utils.files.ByteSource;

import java.time.Instant;

public void beforeEvent(String stepName) {
    ofNullable(Launch.currentLaunch())
            .ifPresent(l -> l.getStepReporter().sendStep(ItemStatus.INFO, stepName));
}

public void afterEvent(boolean passed, byte[] screenshotPng) {
    ofNullable(Launch.currentLaunch()).ifPresent(l -> {
        if (!passed) {
            ReportPortal.emitLog(
                    new ReportPortalMessage(ByteSource.wrap(screenshotPng), "image/png", "Screenshot"),
                    LogLevel.ERROR.name(),
                    Instant.now());
            l.getStepReporter().finishPreviousStep(ItemStatus.FAILED);
        } else {
            l.getStepReporter().finishPreviousStep();
        }
    });
}

Key points of the StepReporter API:

• sendStep(ItemStatus, String) — starts and immediately schedules to finish a virtual step with the given status;
• finishPreviousStep() / finishPreviousStep(ItemStatus) — closes the previous virtual step, optionally overriding its status (useful when the step outcome becomes known only after the next event arrives);
• ReportPortal.emitLog(...) — sends a log / attachment for the currently active test item, whether it is a "real" agent item or a virtual step. Use it for screenshots, page sources, request / response dumps, etc.

Because currentLaunch() works across threads, the same listener can be plugged into any framework without knowing which agent is running underneath, as long as that agent uses this client library (which all official agent-java-* integrations do).

Direct usage: instantiate the client from scratch

If you are writing a custom framework integration, you need to build the client yourself and drive the launch lifecycle end-to-end. The canonical recipe is used by every official agent (see agent-java-junit5 and agent-java-testNG).

Step 1. Build a ReportPortal instance

ReportPortal.builder().build() reads all rp.* parameters from the sources described above (JVM args, environment variables, reportportal.properties). You usually create a single shared instance per JVM:

import com.epam.reportportal.service.ReportPortal;

public static final ReportPortal REPORT_PORTAL = ReportPortal.builder().build();

The builder accepts optional overrides if you need them:

import com.epam.reportportal.listeners.ListenerParameters;
import com.epam.reportportal.service.ReportPortal;
import okhttp3.OkHttpClient;

import java.util.concurrent.Executors;

ListenerParameters params = new ListenerParameters();
params.setBaseUrl("https://reportportal.example.com/");
params.setApiKey("YOUR-API-KEY");
params.setProjectName("my-project");
params.setLaunchName("My Custom Launch");
params.setEnable(true);

ReportPortal rp = ReportPortal.builder()
        .withParameters(params)
        .withHttpClient(new OkHttpClient.Builder())
        .withExecutorService(Executors.newFixedThreadPool(4))
        .build();

If you already have a fully configured low-level ReportPortalClient (for example, when you write tests for the client itself), use the static factory methods ReportPortal.create(client, params) instead of the builder.

Step 2. Start a Launch

A Launch object is a reactive wrapper around the REST API which batches requests, retries failures and finishes in the correct order:

import com.epam.reportportal.service.Launch;
import com.epam.ta.reportportal.ws.model.launch.StartLaunchRQ;
import io.reactivex.Maybe;

import java.time.Instant;

ListenerParameters params = rp.getParameters();

StartLaunchRQ startRq = new StartLaunchRQ();
startRq.setName(params.getLaunchName());
startRq.setDescription(params.getDescription());
startRq.setStartTime(Instant.now());
startRq.setMode(params.getLaunchRunningMode());
startRq.setAttributes(params.getAttributes());
startRq.setRerun(params.isRerun());
if (params.getRerunOf() != null && !params.getRerunOf().isBlank()) {
    startRq.setRerunOf(params.getRerunOf());
}

Launch launch = rp.newLaunch(startRq);
Maybe<String> launchId = launch.start(); // returns a promise; the request is sent asynchronously

Tip: Always register a JVM shutdown hook to finish the launch even when the process is terminated abnormally. Agents do it like this:

Runtime.getRuntime().addShutdownHook(new Thread(() -> {
    FinishExecutionRQ finishRq = new FinishExecutionRQ();
    finishRq.setEndTime(Instant.now());
    launch.finish(finishRq);
}));

If you want to report into an existing launch (for example, a launch started by a primary process in a multi-module build), use ReportPortal.withLaunch(Maybe<String> launchUuid) instead of newLaunch(...). Do not call launch.start() in that case, and do not send finish either — the primary process owns the lifecycle.

Step 3. Start and finish test items

Test items form a tree: suites contain tests, tests contain steps, steps may contain nested steps. Each call returns a Maybe<String> promise with the item UUID which you pass to children and to the finish request:

import com.epam.reportportal.listeners.ItemStatus;
import com.epam.ta.reportportal.ws.model.FinishTestItemRQ;
import com.epam.ta.reportportal.ws.model.StartTestItemRQ;
import io.reactivex.Maybe;

StartTestItemRQ suiteRq = new StartTestItemRQ();
suiteRq.setName("My Suite");
suiteRq.setType("SUITE");
suiteRq.setStartTime(Instant.now());
Maybe<String> suiteId = launch.startTestItem(suiteRq);

StartTestItemRQ stepRq = new StartTestItemRQ();
stepRq.setName("My first test");
stepRq.setType("STEP");
stepRq.setCodeRef("com.example.MyTest.firstTest");
stepRq.setStartTime(Instant.now());
Maybe<String> stepId = launch.startTestItem(suiteId, stepRq);

// ... run the test, report logs, attachments, nested steps ...

FinishTestItemRQ finishStep = new FinishTestItemRQ();
finishStep.setStatus(ItemStatus.PASSED.name());
finishStep.setEndTime(Instant.now());
launch.finishTestItem(stepId, finishStep);

FinishTestItemRQ finishSuite = new FinishTestItemRQ();
finishSuite.setEndTime(Instant.now());
launch.finishTestItem(suiteId, finishSuite);

Allowed item types: SUITE, STORY, TEST, SCENARIO, STEP, BEFORE_CLASS, BEFORE_GROUPS, BEFORE_METHOD, BEFORE_SUITE, BEFORE_TEST, AFTER_CLASS, AFTER_GROUPS, AFTER_METHOD, AFTER_SUITE, AFTER_TEST.

Item statuses are values of the com.epam.reportportal.listeners.ItemStatus enum: PASSED, FAILED, SKIPPED, STOPPED, INTERRUPTED, CANCELLED, INFO, WARN.

Step 4. Send logs and attachments

Logs are attached to the currently active item (determined automatically from the thread that emits them):

import com.epam.reportportal.listeners.LogLevel;
import com.epam.reportportal.message.ReportPortalMessage;
import com.epam.reportportal.service.ReportPortal;
import com.epam.reportportal.utils.files.ByteSource;

import java.time.Instant;

ReportPortal.emitLog("My test started", LogLevel.INFO.name(), Instant.now());

byte[] screenshotPng = takeScreenshot();
ReportPortal.emitLog(
        new ReportPortalMessage(ByteSource.wrap(screenshotPng), "image/png", "Screenshot on failure"),
        LogLevel.ERROR.name(),
        Instant.now());

To log stack traces from caught exceptions use the helper ReportPortal.sendStackTraceToRP(throwable).

Step 5. Finish the launch

Always call launch.finish(...) after the last item is finished — it flushes pending batches, waits for in-flight requests and releases resources:

import com.epam.ta.reportportal.ws.model.FinishExecutionRQ;

FinishExecutionRQ finishRq = new FinishExecutionRQ();
finishRq.setEndTime(Instant.now());
launch.finish(finishRq);

Threading and asynchrony notes

• All startTestItem / finishTestItem / log / emitLog calls are non-blocking — they return immediately and schedule the actual HTTP request on the internal RxJava scheduler. This means exceptions from the server do not propagate back to your code; instead they are logged at WARN / ERROR level, in accordance with the rules described in AGENTS.md. Do not call blockingGet() on the returned Maybe in hot paths.
• ReportPortal.emitLog(...) picks the currently active item from the calling thread. If you emit logs from a worker thread that was not spawned by the agent, make sure the parent item is still open at that moment.
• When rp.reporting.async is false, requests are still asynchronous on the client side but are issued synchronously on the server — useful for tests of the client itself.
• If you need to correlate multiple independent processes into the same launch, enable the multi-process join parameters (see the Multi-process join parameters section above) — no extra code changes are required.

For more elaborate, production-grade examples study the source code of the official agents — they are the best reference for edge cases (retries, reruns, callback reporting, test item tree tracking, etc.):

• agent-java-junit5 — ReportPortalExtension
• agent-java-testNG — TestNGService
• logger-java-selenide — listener that reuses the current Launch