Endtest Bot

An AI-powered autonomous QA colleague that works non-stop in the background to explore your web application and create automated Web Tests.

Endtest Bot is an AI-powered, autonomous QA colleague that can work non-stop in the background to explore a web application, discover useful scenarios, create automated Web Tests, and add completed tests to your Endtest account.

Endtest Bot is currently available as a Beta feature.

Instead of requiring you to describe and create every test individually, Endtest Bot can take a broader QA assignment and autonomously continue exploring, creating, checking, and organizing tests over time.

Unlike the AI Test Creation Agent, which is designed for creating a particular test from your instructions while you are actively working with the test editor, Endtest Bot is designed for longer-running, autonomous AI testing work in the background.

The tests created by Endtest Bot are regular Endtest Web Tests. You can open them, review the steps, edit them, execute them, add them to schedules, or use them in CI/CD just like tests created manually or with the AI Test Creation Agent.

Starting Endtest Bot

To create a new Bot assignment:

  1. Go to the Web Tests section.
  2. Click the Endtest Bot button next to New Test Suite.
  3. Select Generate tests.
  4. Complete the setup wizard.

The Bot will verify that the Starting URL can be opened from the selected Endtest cloud environment before the assignment starts.

If the application is not publicly accessible, see Web Applications with Restricted Access.

1. Choose the web application

Enter the Starting URL for the web application that the Bot should explore.

For example:

https://example.com

The Starting URL must begin with http:// or https://.

You can also open Run configuration on this screen to choose the environment that the Bot should use.

The available settings include:

  • Platform
  • Operating System or Device
  • Browser
  • Browser Version
  • Screen Resolution or Orientation
  • Geolocation

The same run configuration is used while the Bot explores the application and when it checks the tests that it creates.

2. Choose how the Bot should get access

You can choose one of the following access modes.

No credentials required

Use this option when the application or the functionality that you want to test is available without signing in.

Use an existing account

Provide a username or email address and password.

The Bot can use that account to explore authenticated functionality and create tests for signed-in workflows.

For best results, use a dedicated test account instead of a production user account.

Create a new account

Use this option when the application supports first-party signup and you want the Bot to create its own test account.

When possible, the Bot uses a fresh Endtest Mailbox address for the signup flow. This allows it to create tests for scenarios that include email verification.

Creating a new account requires permission to Create data, so that permission is enabled automatically when this access mode is selected.

Endtest Bot does not autonomously sign in through third-party identity providers such as Google, Microsoft, GitHub, Apple, Facebook, or LinkedIn.

3. Choose what the Bot should cover

You can give the Bot a broad assignment or tell it exactly which functionality to automate.

Find useful functionality

The Bot explores the application and identifies useful workflows that can be automated.

This can include navigation, authentication, forms, validation behavior, common user journeys, and other functionality discovered while the Bot builds its QA Map.

Only specific functionality

Use this option when you already know which flows you want automated.

Enter separate flows on separate lines, for example:

Invite a team member
Create a project
Search and filter orders

You can provide up to 12 specific flows in one assignment.

4. Choose the testing types

All testing types are selected by default. You can disable the ones that are not relevant to the assignment.

The available testing types are:

  • Positive testing: happy paths and expected user workflows.
  • Negative testing: invalid inputs, rejected actions, and failure cases.
  • Functional variations: different realistic ways users can perform the main workflow.
  • Boundary & edge cases: minimums, maximums, empty values, and unusual conditions.
  • Validation testing: required fields, formats, constraints, and business-rule validation.
  • Error & recovery testing: failure handling, retries, and recovery states.
  • Accessibility testing: accessibility checks and reports on representative pages.

The initial website exploration still runs even when some testing types are disabled, because it gives the Bot the basic information it needs to understand the application.

5. Set data-change permissions

Endtest Bot can interact with application data, but you control what it is allowed to change.

The available permissions are:

  • Create data: create new accounts, records, or items.
  • Edit data: change existing records.
  • Delete data: remove existing records.

These permissions are hard limits for autonomous Bot work.

For example, if Delete data is disabled, the Bot should not create a test that deletes an existing record simply because it discovers a Delete button during exploration.

Even when the corresponding permissions are enabled, the Bot avoids high-impact actions such as purchases, billing changes, security changes, and sending real external messages.

For applications with important production data, we recommend giving the Bot a dedicated test environment or dedicated test account whenever possible.

6. Add supporting information

You can provide optional Supporting notes with information that may help the Bot understand the application.

For example:

Staging data resets every night.
The Orders page is under Operations.
Use the Demo workspace for all tests.

You can also attach supporting files such as screenshots, PDFs, documents, spreadsheets, CSV files, text files, Markdown files, HTML files, JSON, XML, YAML, or log files.

You can attach up to 4 files, with a maximum of 4 MB per file and 12 MB in total.

Endtest also sends email updates when tests are ready or when the Bot needs input from you.

7. Access check

Before the assignment starts, Endtest launches the selected cloud environment and verifies that the Starting URL is accessible.

If the URL can be opened, the Bot assignment starts automatically.

If the URL cannot be reached, Endtest shows the access problem instead of starting an assignment that cannot make progress.

Applications behind a firewall, VPN, IP allowlist, or other network restriction may require additional configuration. See Web Applications with Restricted Access.

What the Bot does in the background

Once the assignment starts, you can close the Endtest Bot window or leave the Web Tests page.

Endtest Bot is designed to work autonomously and non-stop in the background, without requiring you to keep the Endtest interface open. Its AI explores the application, creates candidate tests, checks them, and gradually builds coverage for the assignment.

Depending on the application and the selected settings, it can create tests for scenarios such as:

  • website and page navigation;
  • positive user workflows;
  • invalid login or signup behavior;
  • first-party signup and email verification;
  • login and logout;
  • form validation;
  • boundary and negative scenarios;
  • requested application workflows;
  • accessibility checks;
  • additional scenarios discovered from the QA Map.

The Bot avoids creating exact duplicate scenarios when equivalent coverage already exists in the same assignment.

Related tests can be grouped into suites such as Authentication, Accessibility, Navigation, Primary Workflows, or Requested Workflows.

For authentication flows, the Bot can also create reusable components so that later tests can reuse common setup such as variables or Login instead of recording the same authentication sequence repeatedly.

How generated tests are checked

Endtest Bot does not publish every generated draft immediately.

For most scenarios, the Bot first creates the test and then runs an internal check to confirm that the generated test can execute successfully before it is added to your visible Web Tests.

If a generated test is not reliable yet, the Bot can retry it later instead of immediately adding an unreliable test to your account.

These internal Bot validation executions are not shown in the normal Results page.

The generated tests themselves are normal Endtest tests and can be executed normally afterward. See How to Execute Web Tests.

Accessibility tests

When Accessibility testing is enabled, Endtest Bot can add Accessibility Check steps to representative pages that it discovers.

Bot-generated Accessibility Check steps use Fail step on severity: Serious by default. This means Serious and Critical accessibility violations can cause the Accessibility Check step to fail, while Moderate and Minor findings remain informational.

An accessibility assertion that finds a Serious or Critical issue is not treated by Endtest Bot as evidence that the generated test itself is broken. The test can still be published because the failed accessibility assertion represents a real finding that the test was designed to detect.

For more information about accessibility checks, see Accessibility Testing.

Bot Activity

Open Endtest Bot → Bot activity to follow current and previous assignments.

Bot Activity shows information such as:

  • the website being tested;
  • the current URL;
  • the task that the Bot is currently working on;
  • recently completed activity;
  • generated tests;
  • tests that need another attempt;
  • requests for user input;
  • links for opening completed tests.

The activity feed also exposes observable browser artifacts when available, including:

  • Screenshot
  • Page source (.txt)

These artifacts can be useful when you want to understand what the Bot saw while it was exploring a page or creating a test.

Older activity entries can be expanded with Show earlier entries.

The activity feed shows observable actions and progress. It does not expose private model reasoning or chain-of-thought.

QA Map

Each Bot assignment can build a QA Map of the application as it explores.

The QA Map can include:

  • public pages;
  • authentication areas;
  • Login and Sign Up branches;
  • authenticated application areas;
  • discovered functionality;
  • planned scenarios;
  • URLs associated with branches;
  • tests that have been linked to those branches.

You can expand or collapse individual branches, or use Expand all and Collapse all.

A test linked to a QA Map branch represents partial coverage, not complete coverage.

For example, a Login branch may eventually contain separate tests for successful login, invalid credentials, required fields, password validation, account-lock behavior, and other scenarios. Creating one Login test does not mean that the entire Login branch is considered fully tested.

The QA Map helps Endtest Bot continue exploring useful gaps without repeatedly creating the same scenario.

Pausing, resuming, and stopping the Bot

An active assignment can be controlled from Bot Activity.

Pause after current task

The Bot finishes the task that it is currently working on and then pauses the assignment.

You can resume it later.

Resume

Continue a paused assignment from its existing progress.

The Bot does not need to restart the assignment from the beginning.

Stop after current task

The Bot finishes the current task and then stops the assignment.

Tests that were already completed remain in your account.

Deleting a Bot Activity history entry does not delete completed tests that were generated by that assignment.

When the Bot needs input

Endtest Bot is designed to continue autonomously whenever its AI can make safe progress, so one blocked scenario does not necessarily stop the rest of the assignment.

If it reaches a scenario where real user input is required, it can mark that work as needing input while continuing with other independent tasks.

A request for input may include the URL, screenshot, and page source that were available when the Bot reached the blocking point.

Endtest also sends an email when the Bot needs input.

CAPTCHA and human verification

Endtest Bot does not solve or bypass reCAPTCHA, hCaptcha, Cloudflare Turnstile, or similar human-verification systems.

If the Bot reaches human verification after completing useful first-party steps, it may preserve those completed steps as partial automation instead of discarding the entire flow.

For testing environments, you can often configure the application or verification provider so trusted Endtest traffic does not receive the challenge.

See Dealing with reCAPTCHA.

Tips for better Bot assignments

For better results:

  1. Use a staging or test environment when the Bot is allowed to create or modify data.
  2. Provide a dedicated test account for authenticated applications.
  3. Use Only specific functionality when you have a small set of high-priority workflows that you want first.
  4. Use supporting notes to explain application-specific terminology, unusual navigation, or test-data requirements.
  5. Enable only the data-change permissions that the Bot actually needs.
  6. Review Bot Activity and the QA Map to understand what has already been explored and where more coverage is being added.
  7. Review generated tests before making them part of a release-blocking CI/CD pipeline, just as you would review tests created by a human or by another AI-assisted workflow.

Endtest Bot vs. AI Test Creation Agent

Both features create regular, editable Endtest Web Tests, but they are designed for different workflows.

Use the AI Test Creation Agent when you want to actively create a particular test from your instructions.

Use Endtest Bot when you want an autonomous AI QA workflow that can work non-stop in the background, exploring the application and creating, checking, and organizing tests over time.

You can use both approaches in the same Endtest project.