Appium Tutorial: A Detailed Guide To Appium Testing

What Changed Between Appium 1, Appium 2, and Appium 3?

Appium 1 bundled every driver in the core. Appium 2 split drivers and plugins into installable packages. Appium 3 requires Node 20.19, drops deprecated endpoints, and ships Inspector as a plugin.

Most Appium content still on the web describes version 1, and that is the single biggest source of broken setups. The table below shows what actually differs across the three releases.

Three of those rows break old scripts silently rather than loudly. Unprefixed capabilities are rejected, TouchAction imports fail to resolve, and a URL ending in wd/hub returns a 404 on Appium 3.

A full breakdown of the newest release is in our guide to Appium 3 features. For upgrading an existing suite, follow the Appium 2 migration guide first, then apply the Node and endpoint changes above.

Note: Appium 3.6.0 shipped in July 2026. Release notes for every version live on the Appium releases page on GitHub.

How Do You Install and Configure Appium?

Install Appium 3 with npm install -g appium, add a platform driver using appium driver install uiautomator2 or xcuitest, then confirm the setup with appium driver doctor before running tests.

Prerequisites are light. You need the Node.js runtime at version 20.19 or newer, npm 10 or newer, and the platform SDK for the operating system you automate.

Install the Appium Server

Install the server globally from npm:

npm install -g appium

Confirm the version before going further:

appium --version

Expected output: a version string such as 3.6.0. A command not found response usually means npm installed to a directory outside your PATH.

Note: Appium Desktop is no longer part of this workflow. That project was archived in 2023, is incompatible with Appium 2 and later, and carries an unpatched remote code execution advisory.

Install a Platform Driver

Since Appium 2, drivers ship separately from the server. A fresh install automates nothing until you add at least one driver.

# See what is available for your platform
appium driver list

# Android
appium driver install uiautomator2

# iOS, macOS hosts only
appium driver install xcuitest

# Confirm what is installed
appium driver list --installed

# Update a driver in place
appium driver update uiautomator2

Verify the Toolchain

Each driver ships its own environment check. Run it before writing a test, because it catches missing SDK paths and signing problems that otherwise surface as confusing session errors.

appium driver doctor uiautomator2

This replaces the standalone appium-doctor package, which is no longer maintained. The background on what those checks cover is in our guide to Appium Doctor.

Install Plugins

Plugins extend the server without forking it. Appium 3 ships Inspector as a plugin, so this is how most teams now get element inspection.

appium plugin list
appium plugin install inspector
appium plugin list --installed

Platform specific setup steps are covered in our walkthrough on how to install Appium.

Run iOS + Android tests written by your AI agent.

Install Appium Skill for Free

How Do You Use Appium Inspector to Find Elements?

Appium Inspector renders the live element tree of a running app, so you can read the exact attributes a locator needs and see why a selector matched nothing before editing the test.

Three ways to run it exist today, and they behave identically once a session starts:

• As the Appium 3 server plugin, installed with appium plugin install inspector.
• As a desktop app from the Appium Inspector releases page for macOS, Linux, and Windows.
• As a hosted web app, which needs no local install but requires a reachable server.

Point it at your server host and port, paste the same capabilities your test will use, then start a session. The app installs on the target device and the element tree loads.

Inspector offers both an accessibility identifier and a generated XPath for the same node. I copy the identifier every time, because the XPath breaks on the next layout change.

More detail on the workflow is in our guide to Appium Inspector for apps.

How Do You Write Your First Appium Test?

Writing an Appium test takes four steps: install the language client, define W3C capabilities in a driver options object, point the client at the running server, then locate elements and assert.

The example below automates an Android app in Python. Before starting, make sure Python 3.9 or newer is installed and a device or emulator is connected.

Step 1: Install the Client Library

pip install Appium-Python-Client

Step 2: Connect a Device

For a physical handset, enable USB debugging under developer options, then confirm the device is visible:

adb devices

Emulators need no cable. Launch one from the device manager in Android Studio instead.

Step 3: Start the Server

appium --port 4723

Expected output: a line reading that the Appium REST http interface listener started on 127.0.0.1:4723. Leave this terminal open while tests run.

Step 4: Write the Test

Capabilities are key value pairs that tell the server which driver, device, and binary to use. Every non standard key needs the appium vendor prefix from version 2 onward.

from appium import webdriver
from appium.options.android import UiAutomator2Options
from appium.webdriver.common.appiumby import AppiumBy
from selenium.webdriver.support.ui import WebDriverWait
from selenium.webdriver.support import expected_conditions as EC

capabilities = {
    "platformName": "Android",
    "appium:platformVersion": "14",
    "appium:deviceName": "Pixel 8",
    "appium:app": "/apk/com.slot.spin.game.play.apk",
    "appium:appPackage": "com.slot.spin.game.play",
    "appium:appActivity": ".MainActivity",
    "appium:automationName": "UiAutomator2",
}

options = UiAutomator2Options().load_capabilities(capabilities)

# Appium 2 and 3 serve the W3C endpoint at the root. A trailing
# /wd/hub returns 404 on these versions.
driver = webdriver.Remote("http://127.0.0.1:4723", options=options)

try:
    # An explicit wait returns the moment the element is ready,
    # so the test stays fast on quick devices and stable on slow ones.
    spin_button = WebDriverWait(driver, 20).until(
        EC.element_to_be_clickable(
            (AppiumBy.ID, "com.slot.spin.game.play:id/playBtn")
        )
    )
    spin_button.click()

    result = WebDriverWait(driver, 20).until(
        EC.presence_of_element_located(
            (AppiumBy.ID, "com.slot.spin.game.play:id/resultText")
        )
    )
    print("Spin result:", result.text)
finally:
    driver.quit()

Expected output: the app launches on the device, the spin button is tapped, and the terminal prints a line beginning with Spin result followed by the on screen value.

Java teams use the same capability keys through a typed options object. The client differs, the protocol underneath does not.

UiAutomator2Options options = new UiAutomator2Options()
        .setPlatformVersion("14")
        .setDeviceName("Pixel 8")
        .setApp("/apk/com.slot.spin.game.play.apk")
        .setAppPackage("com.slot.spin.game.play")
        .setAppActivity(".MainActivity");

AndroidDriver driver = new AndroidDriver(
        new URL("http://127.0.0.1:4723"), options);

WebDriverWait wait = new WebDriverWait(driver, Duration.ofSeconds(20));
wait.until(ExpectedConditions.elementToBeClickable(
        AppiumBy.id("com.slot.spin.game.play:id/playBtn"))).click();

driver.quit();

Building capability sets by hand gets tedious across a device matrix. Our capabilities generator produces valid W3C blocks for either client.

Note: Ready to run this suite against physical hardware? Start free on 10,000+ real devices.

Which Locator Strategy Should You Use in Appium?

Prefer accessibility id in Appium because it maps to content-desc on Android and the accessibility identifier on iOS. Fall back to resource id, then class chain, and avoid XPath in hot paths.

Locator choice is the single largest cause of flaky mobile suites, and nearly every unstable suite I have inherited had XPath at its core. Those queries force the driver to serialise the whole element tree.

Getting to priority one is a development task, not a QA task. Someone has to set content-desc on Android views and accessibilityIdentifier on iOS views, ideally to the same string.

When that groundwork exists, one locator serves both platforms:

# Same locator, both platforms, no branching
driver.find_element(AppiumBy.ACCESSIBILITY_ID, "checkout_button")

# Android only, still fast
driver.find_element(AppiumBy.ID, "com.shop.app:id/checkout")

# iOS class chain, prefer over XPath
driver.find_element(
    AppiumBy.IOS_CLASS_CHAIN,
    '**/XCUIElementTypeButton[`label == "Checkout"`]'
)

A common failure looks like a locator problem but is not. If an element is present in Inspector yet invisible to the test, check whether the app switched into a webview context.

Our guide to locators in Appium holds a wider catalogue of selectors and the syntax each one expects.

How Do You Handle Waits in Appium?

Use WebDriverWait with expected conditions instead of fixed sleeps. An explicit wait returns as soon as the element is ready, so tests stay fast on quick devices and stable on slow ones.

Fixed sleeps fail in both directions. Too short and the suite breaks on a cold device. Too long and every run pays the worst case, which compounds across hundreds of steps.

Mixing implicit and explicit waits is the trap. I lost most of an afternoon to a five second wait that blocked for thirty, because a leftover implicit wait was compounding with it.

# Pick one strategy. If you use explicit waits, leave
# the implicit wait at zero.
driver.implicitly_wait(0)

wait = WebDriverWait(driver, 20)

# Present in the tree, may still be off screen
wait.until(EC.presence_of_element_located(
    (AppiumBy.ACCESSIBILITY_ID, "cart_badge")))

# Rendered and tappable, the condition most steps want
wait.until(EC.element_to_be_clickable(
    (AppiumBy.ACCESSIBILITY_ID, "checkout_button"))).click()

# Gone, useful after dismissing a loader
wait.until(EC.invisibility_of_element_located(
    (AppiumBy.ACCESSIBILITY_ID, "loading_spinner")))

Animations cause a subtle version of this. An element can be clickable while still sliding into place, so the tap lands on the wrong coordinates and the step fails intermittently.

Waiting on a post animation condition, such as a settled label, is more reliable than waiting on the element itself. Gesture level timing is covered in Appium gestures.

How Do You Run Appium Tests on Real Devices at Scale?

Point the same test at a device cloud instead of localhost. Only the remote URL and a few vendor prefixed capabilities change, so one suite fans out across many real Android and iOS handsets.

A local setup proves the test works. It does not prove the app works across the device and OS spread your users actually run, and maintaining that hardware is a full time job.

TestMu AI is an AI native test orchestration and execution platform whose real device cloud runs Appium suites on 10,000+ Android and iOS handsets.

Your test logic stays untouched. Everything below the assertions is the same suite you wrote above, now pointed somewhere else.

• Real device grid: run against shipped hardware, not emulators, on demand.
• Parallel execution: fan one suite across many device and OS combinations at once.
• Device logs and video: every session records logs and a replay for triage.
• Network simulation: reproduce throttled and offline conditions that break real users.
• Geolocation testing: validate location gated flows without physically travelling.

Setup steps and capability reference live in the getting started with Appium testing documentation.

Only two things change in the script. The remote URL points at the cloud hub, and every vendor setting moves into an lt:options block beside the standard capabilities.

from appium import webdriver
from appium.options.android import UiAutomator2Options

username = "YOUR_USERNAME"
access_key = "YOUR_ACCESS_KEY"

capabilities = {
    "platformName": "Android",
    "appium:platformVersion": "14",
    "appium:deviceName": "Pixel 8",
    "appium:app": "lt://APP_ID",
    "appium:automationName": "UiAutomator2",
    # Cloud vendor settings live in lt:options, not the appium namespace
    "lt:options": {
        "username": username,
        "accessKey": access_key,
        "build": "Slot Machine Regression",
        "name": "Spin flow, Pixel 8",
        "isRealMobile": True,
        "network": True,
        "video": True,
        "deviceLog": True,
    },
}

options = UiAutomator2Options().load_capabilities(capabilities)

remote_url = f"https://{username}:{access_key}@mobile-hub.lambdatest.com/wd/hub"
driver = webdriver.Remote(remote_url, options=options)

# The rest of the test is identical to the local version.

One difference is worth planning for. The app capability takes an lt:// identifier returned by the upload API, not a local file path.

That means a build upload step belongs in the pipeline before test execution starts.

Fanning the same suite across devices is where the time is recovered. Our guide to Appium Parallel Testing covers session limits and result aggregation.

How Do You Fix Common Appium Errors?

Most Appium errors trace to environment drift rather than test logic. Match the message to its cause below, then rerun the driver doctor check before editing any locators or waits.

The table maps each message practitioners actually see to the cause behind it.

Two of those rows deserve emphasis because they mimic real bugs. A session that dies mid run looks like an app crash, and a context mismatch looks like a missing element.

Now I rerun the driver doctor check before reading a single stack trace. Toolchain drift after an Xcode or Android SDK upgrade explains more failures than any change your team made to the test cases.

Server logs are the other underused signal. Running the server with the debug log level prints the exact capability payload it received, which settles most argument mismatches in seconds.

When Should You Not Use Appium?

Skip Appium when you need frame accurate gaming input, deep OS level control, or sub second unit feedback. Espresso and XCUITest run in process and give faster, more stable results there.

The cost of the cross platform abstraction is real. Every command crosses HTTP and a driver boundary, which adds latency and one more layer that can fail independently of your app.

Four constraints show up often enough to plan around:

• iOS automation requires a macOS host, which constrains CI runner choice.
• Flutter apps render to a canvas, so standard drivers see no element tree.
• Reporting is minimal, so test reports come from your runner.
• Every iOS CI job pays the WebDriverAgent build again, which inflates pipeline time.

A pragmatic split works well in practice. Developers keep fast in process tests per platform, and QA owns one Appium suite covering the cross platform journeys that matter to release decisions.

Full evaluations of each option are in our roundup of Appium alternatives.

Key Takeaways

Appium earns its place when one app ships to two platforms and you refuse to modify the binary to test it. That constraint, not raw speed, is what the framework optimises for.

Three decisions determine whether a suite stays healthy. Install drivers explicitly, standardise on accessibility identifiers, and replace every fixed sleep with an explicit wait before the suite grows.

Next step: run the test above on a real handset, then keep the Appium commands cheat sheet open while you extend it. Teams formalising skills can take the Appium 101 certification.

Verified against Appium 3.6.0 and the official Appium documentation in August 2026.