Skip to content

Self-Driving Tests with Explorbot

Explorbot Explorbot

Writing end-to-end tests takes time, and most apps have more pages than anyone has time to cover. Explorbot is an AI agent that does that work for you. Point it at a URL, and it explores the app, plans scenarios, runs them in a browser, and writes the passing flows down as CodeceptJS tests.

You don’t write a script or supervise the run. That’s what we mean by a self-driving test.

Explorbot is built by the CodeceptJS team at Testomat.io and runs on CodeceptJS under the hood.

How a session works

Section titled “How a session works”

Give Explorbot a URL, and a crew of agents takes it from there:

  1. Research. Map the page into sections and index every element. No source code or docs needed.
  2. Plan. Draft test scenarios: happy paths, curious detours, and edge cases.
  3. Execute. Drive the browser step by step, adapting as the app changes.
  4. Verify. Confirm each outcome, group findings by root cause, and capture evidence.
  5. Keep. Save passing flows as tests, with reports and screencasts. What it learned is reused on the next run.

The workflow itself (research, plan, test) is fixed and predictable. AI makes the small decisions: which button to click, how to close a modal, how to recover from an error. Most of the clicking and reading runs on fast, cheap models, so a continuous run costs about $1 per hour.

Try it

Section titled “Try it”

You need Node.js 24+ (or Bun) and an API key from an AI provider such as OpenRouter, Groq, or OpenAI.

Terminal window
npm i explorbot --save
npx playwright install
npx explorbot init

Add your provider key to .env, set your app URL in explorbot.config.js, and start on a focused section of the app, such as an admin panel or a settings page:

Terminal window
npx explorbot start /admin/users

Type /explore, and Explorbot runs its loop on its own.

What you get: CodeceptJS tests

Section titled “What you get: CodeceptJS tests”

Every plan Explorbot runs ends up as a regular CodeceptJS test file in output/tests/. CodeceptJS is the default output:

import step, { Section } from 'codeceptjs/steps';


Feature('Runs Archive Feature Testing')


Before(({ I }) => {
  I.amOnPage('/projects/zyntra/runs/archive');
  I.wait(1);
});


Scenario('Apply filters specific to archived runs and verify results', ({ I }) => {
  Section('Open the filter panel');
  I.click({ css: 'button.btn-only-icon.btn-lg:has(svg.md-icon-filter)' });


  Section('Pick the Passed status');
  I.click('Select status');
  I.click('Passed');
  I.click('Apply');


  I.see('1 run found');
});

The locators are the ones Explorbot actually used during the session, and each Section is labeled with what that group of steps does. The file reads top to bottom like the plan it came from.

The file always runs, even when some scenarios failed:

  • Passed scenarios become Scenario(...).
  • Failed scenarios become Scenario.skip(...) with a // FAILED: comment, so you can see what broke.
  • Scenarios Explorbot didn’t reach become Scenario.todo(...), with the planned steps kept as comments.

Commit it as-is, or refactor it into page objects like any other test. To replay it with self-healing of broken steps, run:

Terminal window
npx explorbot rerun output/tests/runs_archive_feature_testing.js

Next to your existing suite

Section titled “Next to your existing suite”

Explorbot doesn’t replace your regression tests. It covers what they don’t reach. Your CodeceptJS suite replays the same steps every build. Explorbot explores the same pages in new ways each run, clicking UI and paths your scripts never touch. Point it at a new feature with no tests, and it works out the basic test cases and runs them right away.

If your CI runs Playwright, it can run Explorbot. A nightly job with a test budget re-explores the app while nobody watches, and the results can be sent to Testomat.io with screencasts of every test.

Learn more

Section titled “Learn more”