end-to-end teszt (E2E)

in English: end-to-end test (E2E test)

advancedwidespread

A software test that tries out the program as a real user would: from the first click to the final result, including every part.

In plain terms

Imagine you are opening a restaurant. You can taste the soup separately, try out the cash register, and check if the oven works. This still does not tell you what happens when a guest enters: are they seated, does the order reach the kitchen, is the food brought out, and can the bill be paid? The end-to-end test is this trial dinner: it goes the whole way from the door to the receipt.

In software, this means that a program automatically opens the application, clicks, types, fills out a form, and checks whether what was supposed to happen actually happened at the end. Meanwhile, the entire system is working: the interface visible on the screen, the server running in the background, the database, and often external services, such as payment.

It is needed because a significant share of errors arise where the individual parts meet, even though each works well on its own. Smaller tests examine the components, while the E2E test checks whether the whole machine starts up.

In the era of coding agents, testing has become even more critical. When an AI agent writes or modifies code, running tests is one of the most useful signals that key user paths still function after changes. However, this does not prove that nothing broke, as tests only verify the paths that were written.

How it works

The team selects the most critical user paths, such as registration, login, purchase, or transfer.
These are described step by step in a test program: open the page, enter this, click that, and this must appear.
The test runs in a realistic but isolated sandbox environment with test data to avoid affecting real customers or money.
An automated tool controls a real browser or mobile app, executes the steps, and compares the result with the expected outcome at every point.
If something differs, the test fails and saves a screenshot or log of the error. Teams typically run tests automatically: the most important ones after code changes, and the full suite, for example, overnight and before a release.

An everyday example

At an online store, a test program "shops" every night and before every release: it adds a product to the cart, enters a shipping address, pays with a test card, and checks if a confirmation email arrived. If a modification causes the payment button to stop responding, this is discovered before you, as a customer, encounter it.

Who is affected

developerThey write and maintain these tests, and the tests alert them if a change or the work of a coding agent has broken something elsewhere. → Code

entrepreneur, executiveErrors in revenue-generating processes, such as ordering and payment, cause direct losses that this test can catch before launch.

everyoneIf your online banking, ticketing site, or service application continues to work after an update, such tests are often behind it.

A common misconception

MythIf all end-to-end tests are green, then the program is flawless.

In realityThe test only checks the paths that were pre-written for it. Errors in rare situations, unusual data, and untested features may still remain in the program.

Limits and risks

Slow and expensive: running a full sequence can take minutes or hours, so only the most critical paths are worth verifying this way.
Flaky: it fails intermittently. A slow network or a timing issue can make it fail even when the program has no bug, which undermines trust in the tests.
It requires much maintenance because even minor interface changes, such as renaming or moving a button, can break the test.
It shows that something is wrong but rarely where exactly: the root cause must be found using smaller, targeted tests.

Similar, but not the same

unit test – Unit testing examines a single small part of a program in isolation, while end-to-end testing examines the entire system as a whole from the user's perspective.
integration test – Integration testing checks the cooperation of several connected components but does not trace the entire path from screen to database.
manual testing – Manual testing is performed by a human clicking through, while end-to-end testing is typically automated, making it repeatable at any time in the same way.

Related concepts

Broader

software testing, quality assurance, test automation

Narrower

browser test, user interface test

On the edge

unit test, integration test, regression test, smoke test, tesztpiramis, continuous integration

A brief history

Testing the entire system as a whole is an old software testing practice; the automated, browser-controlling variant spread with web applications. One milestone was the Selenium tool launched in 2004.

Where things stand

At the end of 2026, it is a standard part of serious software development, built into the automated checks that run before release. It is increasingly common for AI agents to write, run and fix the tests, and for coding agents to use them to check their own work.

Status: 2026 November

Also known as

end-to-end teszt · E2E teszt · E2E · end-to-end testing · end-to-end test · end-to-end testing · end-to-end testing · E2E test

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.