The biggest change in Selenium recently has been the inclusion of the WebDriver API. Driving a browser natively as a user would either locally or on a remote machine using the Selenium server, it marks a leap forward in terms of browser automation.

Selenium WebDriver fits in the same role as RC did, and has incorporated the original 1.x bindings. It refers to both the language bindings and the implementations of the individual browser controlling code. This is commonly referred to as just WebDriver or sometimes as Selenium 2.

Selenium 1.0 + WebDriver = Selenium 2.0

Driver requirements

Through WebDriver, Selenium supports all major browsers on the market such as Chrom(ium), Firefox, Internet Explorer, Opera, and Safari. Where possible, WebDriver drives the browser using the browser's built-in support for automation, although not all browsers have official support for remote control.

WebDriver's aim is to emulate a real user's interaction with the browser as closely as possible. This is possible at varying levels in different browsers. For more details on the different driver idiosyncracies, please see Driver Idiosyncracies.

Even though all the drivers share a single user-facing interface for contolling the browser, they have slightly different ways of setting up browser sessions. Since many of the driver implementations are provided by third parties, they are not included in the standard Selenium distribution.

Driver instantiation, profile management, and various browser specific settings are examples of parameters that have different requirements depending on the browser. This section explains the basic requirements for getting you started with the different browsers.

Adding Executables to your PATH

Most drivers require an extra executable for Selenium to communicate with the browser. You can manually specify where the executable lives before starting WebDriver, but this can make your tests less portable, as the executables will need to be in the same place on every machine, or included within your test code repository.

By adding a folder containing WebDriver's binaries to your system's path, Selenium will be able to locate the additional binaries without requiring your test code to locate the exact location of the driver.

Quick reference

Browser Supported OS Maintained by Download Issue Tracker
Chromium/Chrome Windows
Google Downloads Issues
Firefox Windows
Mozilla Downloads Issues
Edge Windows 10 Microsoft Downloads Issues
Internet Explorer Windows Selenium Project Downloads Issues
Safari macOS El Capitan and newer Apple Built in Issues
Opera Windows
Opera Downloads Issues


To drive Chrome or Chromium, you have to download chromedriver and put it in a folder that is on your system's path.

On Linux or macOS, this means modifying the PATH environmental variable. You can see what directories, separated by a colon, make up your system's path by executing the following command:

$ echo $PATH

To include chromedriver on the path if it isn't already, make sure you include the chromedriver binary's parent directory. The following line will set the PATH environmental variable its current content, plus an additional path added after the colon:

$ export PATH="$PATH:/path/to/chromedriver"

When chromedriver is available on your path, you should be able to execute the _chromedriver_ executable from any directory.

To instantiate a Chrome/Chromium session, you can do the following:

WebDriver driver = new ChromeDriver();

Remember that you have to set the path to the chromedriver executable. This is possible using the following line:

System.setProperty("webdriver.chrome.driver", "/path/to/chromedriver");
require "selenium-webdriver"
driver = Selenium::WebDriver.for :chrome

The chromedriver is implemented as a WebDriver remote server that by exposing Chrome's internal automation proxy interface instructs the browser what to do.


Starting with Selenium 3, Mozilla has taken over implementation of Firefox Driver, with geckodriver. The new driver for Firefox is called geckodriver and works with Firefox 48 and newer. Since the Firefox WebDriver is under development, the newer the Firefox version the better the support.

As geckodriver is the new default way of launching Firefox, you can instantiate Firefox in the same way as Selenium 2:

WebDriver driver = new FirefoxDriver();
require "selenium-webdriver"
driver = Selenium::WebDriver.for :firefox

If you prefer not to set geckodriver's location using PATH, set the geckodriver binary location programmatically:

System.setProperty("webdriver.gecko.driver", "/path/to/geckodriver");

It is also possible to set the property at run time:

mvn test -Dwebdriver.gecko.driver=/path/to/geckodriver

It is currently possible to revert to the older, more feature complete Firefox driver, by installing Firefox 47.0.1 or 45 ESR and specifying a desired capability of marionette as false. Later releases of Firefox are no longer compatible.


Edge is Microsoft's newest browser, included with Windows 10 and Server 2016. Updates to Edge are bundled with major Windows updates, so you'll need to download a binary which matches the build number of your currently installed build of Windows. The Edge Developer site contains links to all the available binaries. Bugs against the EdgeDriver implementation can be raised with Microsoft. If you'd like to run tests against Edge, but aren't running Windows 10, Microsoft offer free VMs for testers on the Edge Developer site.

WebDriver driver = new EdgeDriver();

If Edge driver is not present in your path, you can set the path using the following line:

System.setProperty("webdriver.edge.driver", "/path/to/edgedriver");

Internet Explorer

Internet Explorer was Microsoft's default browser until Windows 10, although it is still included in Windows 10. Internet Explorer Driver is the only driver The Selenium project aims to support the same releases Microsoft considers current. Older releases may work, but will be unsupported.

While the Selenium project provides binaries for both the 32-bit and 64-bit versions of Internet Explorer, there are some limitations with Internet Explorer 10 & 11 with the 64-bit driver, but using the 32-bit driver continues to work well. It should be noted that as Internet Explorer preferences are saved against the logged in user's account, some additional setup is required.

WebDriver driver = new InternetExplorerDriver();

If Internet Explorer driver is not present in your path, you can set the path using the following line:

System.setProperty("webdriver.ie.driver", "/path/to/iedriver");
Microsoft also offer a WebDriver binary for Internet Explorer 11 on Windows 7 & 8.1. It has not been updated since 2014 and is based of a draft version of the W3 specification. Jim Evans has an excellent writeup on Microsoft's implementation.

Opera (Presto-based)

Setting up support for Opera is fairly similar to the Chrome approach in that they are both supported by third parties. Consequently, you need to set up a dependency to the com.opera:operadriver artifact (if using Maven) or download a prebuilt standalone server JAR. As long as the JAR is available on your classpath, you should be good to go.

If you use Python or Ruby as your language of choice, you must set the SELENIUM_SERVER_JAR environmental variable to point to the location of the binary JAR.

On Linux and macOS, you can do this for your current shell session by running:

$ export SELENIUM_SERVER_JAR=/path/to/operadriver.jar

It's possible to set this in your shell's profile (e.g. ~/.bashrc) to have the variable set every time you launch a new shell session. Alternatively, you can set the variable programmatically in your test runner. In Python, you'd do something along the lines of

import os
os.environ["SELENIUM_SERVER_JAR"] = "/path/to/selenium-server-standalone.jar"

The equivalent in Ruby:

ENV["SELENIUM_SERVER_JAR"] = "/path/to/selenium-server-standalone.jar"

Instantiating a driver session is similar to Firefox and Chromium:

WebDriver driver = new OperaDriver();
require "selenium-webdriver"
driver = Selenium::WebDriver.for :opera


Starting with Safari 10 on macOS El Capitan and Sierra, WebDriver support is included with each release of the browser. To enable support:

  1. Enable the Developer menu from Safari preferences
  2. Check the Allow Remote Automation option from with the Develop menu
  3. Run
    /usr/bin/safaridriver -p 1337
    from the terminal for the first time and type your password at the prompt to authorise WebDriver

You can then start a driver session using:

WebDriver driver = new SafariDriver();

Those looking to automate Safari on iOS should look to the Appium project. Whilst Safari was previously available for Windows, Apple has long since dropped support, making it a poor choice of test platform.

Mock browsers



PhantomJS is a headless browser based on Webkit, albeit a version much older than that used by Google Chrome or Safari. Unlike HtmlUnit driver, it also supports Javascript. Whilst historically a popular choice, it would now be wise to avoid PhantomJS. The project has been unmaintained since the 5th of August, so whilst the web will continue to change, PhantomJS will not be updated. This was after Google announced the ability to run Chrome headlessly, something also now offered by Mozilla's Firefox.

Browser launching and manipulation


Ruby is not installed by default on Windows. Download the latest version and run the installer. You can leave all settings at default values, except at the Installation Destination and Optional Tasks screen check Add Ruby executables to your PATH checkbox. To drive any browser, you have to install selenium-webdriver Ruby gem. To install it, open command prompt and type this:

$ gem install selenium-webdriver

Internet Explorer

Internet Explorer is installed by default on Windows, so no installation is needed. To drive Internet Explorer on Windows, you have to download the latest Internet Explorer Driver and put the file into a folder that is in PATH. To find out which directories are in PATH, type echo %PATH% in command prompt.

$ echo %PATH%
C:\Ruby200\bin looks like a good place. Unzip `IEDriverServer` file and move `IEDriverServer.exe` there. This should open a new Internet Explorer window:

System.setProperty("webdriver.ie.driver", /path/to/internetexplorerdriver);
WebDriver driver = new InternetExplorerDriver();
require "selenium-webdriver"
driver = Selenium::WebDriver.for :internet_explorer


WebDriver can generally be said to have a blocking API. Because it is an out-of-process library that instructs the browser what to do, and because the web platform has an intrinsically asynchronous nature, WebDriver doesn't track the active, real-time state of the DOM. This comes with some challenges that we will discuss here.

From experience, most intermittents that arise from use of Selenium and WebDriver are connected to race conditions that occur between the browser and the user's instructions. An example could be that the user instructs the browser to navigate to a page, then gets a no such element error when trying to find an element.

Consider the following document:

<!doctype html>
<meta charset=utf-8>
<title>Race Condition Example<title>

  var initialised = false;
  window.addEventListener("load", function() {
    var newElement = document.createElement("p");
    newElement.textContent = "Hello from JavaScript!";
    initialised = true;

The WebDriver instructions might look innocent enough:

el = driver.find_element_by_tag_name("p")
assert el.text == "Hello from JavaScript!"

The issue here is that the default page load strategy used in WebDriver listens for the document.readyState to change to "complete" before returning from the call to navigate. Because the p element is added after the document has completed loading, this WebDriver script might be intermittent. It “might” be intermittent because no guarantees can be made about elements or events that trigger asynchronously without explicitly waiting—or blocking—on those events.

Fortunately, using the normal instruction set available on the WebElement interface—such as WebElement.click and WebElement.sendKeys—are guaranteed to be synchronous, in that the function calls won't return (or the callback won't trigger in callback-style languages) until the command has been completed in the browser. The advanced user interaction APIs, Keyboard and Mouse, are exceptions as they are explicitly intended as “do what I say” asynchronous commands.

Waiting is having the automated task execution elapse a certain amount of time before continuing with the next step.

To overcome the problem of race conditions between the browser and your WebDriver script, most Selenium clients ship with a wait package. When employing a wait, you are using what is commonly referred to as an explicit wait.

Explicit waiting

Explicit waits are available to Selenium clients for imperative, procedural languages. They allow your code to halt program execution, or freeze the thread, until the condition you pass it resolves. The condition is called with a certain frequency until the timeout of the wait is elapsed. This means that for as long as the condition returns a falsy value, it will keep trying and waiting.

Since explicit waits allow you to wait for a condition to occur, they make a good fit for synchronising the state between the browser and its DOM, and your WebDriver script.

To remedy our buggy instruction set from earlier, we could employ a wait to have the findElement call wait until the dynamically added element from the script has been added to the DOM:

from selenium.webdriver.support.ui import WebDriverWait

def document_initialised(driver):
    return driver.execute_script("return initialised")

el = driver.find_element_by_tag_name("p")
assert el.text == "Hello from JavaScript!"

We pass in the condition as a function reference that the wait will run repeatedly until its return value is truthy. A “truthful” return value is anything that evaluates to boolean true in the language at hand, such as a string, number, a boolean, an object (including a WebElement), or a populated (non-empty) sequence or list. That means an empty list evaluates to false. When the condition is truthful and the blocking wait is aborted, the return value from the condition becomes the return value of the wait.

With this knowledge, and because the wait utility ignores no such element errors by default, we can refactor our instructions to be more concise:

from selenium.webdriver.support.ui import WebDriverWait

el = WebDriverWait(driver).until(lambda d: return d.find_element_by_tag_name("p"))
assert el.text == "Hello from JavaScript!"

In that example, we pass in an anonymous function (but we could also define it explicitly as we did earlier so it may be reused). The first and only argument that is passed to our condition is always a reference to our driver object, WebDriver (called d in the example). In a multi-threaded environment, you should be careful to operate on the driver reference passed in to the condition rather than the reference to the driver in the outer scope.

Because the wait will swallow no such element errors that are raised when the element isn't found, the condition will retry until the element is found. Then it will take the return value, a WebElement, and pass it back through to our script.

If the condition fails, e.g. a truthful return value from the condition is never reached, the wait will throw/raise an error/exception called a timeout error.


The wait condition can be customised to match your needs. Sometimes it's unnecessary to wait the full extent of the default timeout, as the penalty for not hitting a successful condition can be expensive.

The wait lets you pass in an argument to override the timeout:

WebDriverWait(driver, timeout=3).until(some_condition)

Expected conditions

Because it's quite a common occurrence to have to synchronise the DOM and your instructions, most clients also come with a set of predefined expected conditions. As might be obvious by the name, they are conditions that are predefined for frequent wait operations.

The conditions available in the different language bindings vary, but this is a non-exhaustive list of a few:

alert is present
element exists
element is visible
title contains
title is
element staleness
visible text

You can refer to the API documentation for each client binding to find an exhaustive list of expected conditions:

Implicit waiting

There is a second type of wait that is distinct from explicit waits called implicit waiting. By implicitly waiting, WebDriver polls the DOM for a certain duration when trying to find any element. This can be useful when certain elements on the webpage are not available immediately and need some time to load.

Implicit waiting for elements to appear is disabled by default and will need to be manually enabled on a per-session basis. Mixing explicit waits and implicit waiting will cause unintended consequences, namely waits sleeping for the maximum time even if the element is available or condition is true.

Warning: Do not mix implicit and explicit waits. Doing so can cause unpredictable wait times. For example, setting an implicit wait of 10 seconds and an explicit wait of 15 seconds could cause a timeout to occur after 20 seconds.

An implicit wait is to tell WebDriver to poll the DOM for a certain amount of time when trying to find an element or elements if they are not immediately available. The default setting is 0, meaning disabled. Once set, the implicit wait is set for the life of the session.

WebDriver driver = new FirefoxDriver();
driver.manage().timeouts().implicitlyWait(10, TimeUnit.SECONDS);
WebElement myDynamicElement = driver.findElement(By.id("myDynamicElement"));


FluentWait instance defines the maximum amount of time to wait for a condition, as well as the frequency with which to check the condition.

Users may configure the wait to ignore specific types of exceptions whilst waiting, such as NoSuchElementExceptions when searching for an element on the page.

// Waiting 30 seconds for an element to be present on the page, checking
// for its presence once every 5 seconds.
Wait<WebDriver> wait = new FluentWait<WebDriver>(driver)
  .withTimeout(30, SECONDS)
  .pollingEvery(5, SECONDS)

WebElement foo = wait.until(new Function<WebDriver, WebElement>() {
  public WebElement apply(WebDriver driver) {
    return driver.findElement(By.id("foo"));
FluentWait<By> fluentWait = new FluentWait<By>(By.tagName("TEXTAREA"));
fluentWait.pollingEvery(100, TimeUnit.MILLISECONDS);
fluentWait.withTimeout(1000, TimeUnit.MILLISECONDS);
fluentWait.until(new Predicate<By>() {
  public boolean apply(By by) {
    try {
      return browser.findElement(by).isDisplayed();
    } catch (NoSuchElementException ex) {
      return false;
browser.findElement(By.tagName("TEXTAREA")).sendKeys("text to enter");

Support classes

HTTP proxies

Page loading strategy

Web elements