Symfony Panther, Playwright/ChromeDriver for PHP

If your stack is PHP, you have three options for browser automation: shell out to a Python script, embed a headless tool over CDP, or use Symfony Panther. Panther is the cleanest of the three, a native PHP library that drives Chromium or Firefox via WebDriver, with an API that feels like Goutte/DomCrawler but executes real JavaScript.

Install

composer require symfony/panther

Panther can use either Chromium (via ChromeDriver) or Firefox (via geckodriver). The driver binary is downloaded automatically the first time you run a script that needs it, into your project's drivers/ folder.

Symfony users get extra wiring (test framework integration, Symfony Console support). Non-Symfony PHP works fine too, Panther is composer require-able from any project.

Your first scraper

<?php
require_once __DIR__ . '/vendor/autoload.php';

use Symfony\Component\Panther\Client;

$client = Client::createChromeClient();
$crawler = $client->request('GET', 'https://practice.scrapingcentral.com/');

echo $crawler->filter('h1')->first()->text() . PHP_EOL;

$client->quit();

Client::createChromeClient() launches headless Chromium. $client->request() is the navigation method (note the HTTP-verb-style API, same as Symfony's BrowserKit). The returned $crawler is a DomCrawler instance, the same one you've used for static scraping in Sub-Path 1.

For Firefox: Client::createFirefoxClient().

What Panther actually does

Under the hood, Panther:

1. Spawns ChromeDriver (or geckodriver) as a subprocess.
2. Speaks the WebDriver protocol to it (the same protocol Selenium uses).
3. Exposes a PHP API that mimics Symfony's BrowserKit\Client and DomCrawler\Crawler.

Result: you write PHP that looks like a Symfony Crawler scrape, but the page is being driven by a real browser that runs JS, executes XHRs, and updates the DOM. Same conceptual model as Playwright, different transport.

Crawler + WebDriver

Two APIs are available on every Panther response:

$crawler = $client->request('GET', 'https://practice.scrapingcentral.com/products');

// DomCrawler-style (read-only, familiar from Sub-Path 1)
foreach ($crawler->filter('.product-card') as $node) {
  echo $node->getElementsByTagName('h2')[0]->textContent . PHP_EOL;
}

// WebDriver-style (interactive)
$client->getWebDriver()->findElement(WebDriverBy::cssSelector('button.load-more'))->click();

For pure extraction, stay in the Crawler API, it's the PHP idiom. For interactions (click, fill, scroll), reach for WebDriver. Panther blends them seamlessly.

Waiting for JS-rendered content

use Symfony\Component\Panther\DomCrawler\Crawler;

$client = Client::createChromeClient();
$crawler = $client->request('GET', 'https://practice.scrapingcentral.com/challenges/dynamic/spa-pure');

// Wait up to 10 seconds for the .product-grid to appear
$client->waitFor('.product-grid', 10);

// Now extract
$cards = $crawler->filter('.product-card');
echo "Found " . $cards->count() . " products" . PHP_EOL;

$client->quit();

waitFor(selector, timeout) polls until the selector resolves to at least one element. Variants:

$client->waitForVisibility('.modal');
$client->waitForInvisibility('.spinner');
$client->waitForElementToContain('.status', 'Loaded');
$client->waitForAttributeToContain('input', 'value', 'success');

These are Panther's equivalents to Playwright's wait_for_selector(state=...). Same idea, different API surface.

Filling forms and clicking

$client = Client::createChromeClient();
$crawler = $client->request('GET', 'https://practice.scrapingcentral.com/account/login');

$form = $crawler->filter('form')->form([
  'email'  => 'demo@example.com',
  'password' => 'password',
]);
$client->submit($form);

$client->waitFor('.welcome-message');
echo $crawler->filter('.welcome-message')->text();

$client->quit();

The form API is straight from Symfony BrowserKit, the same code that works against a static crawler also works against a JS-driven page, with the addition of waitFor to handle async post-submit redirects.

For more direct control:

$crawler->selectButton('Login')->click();
$crawler->filter('input[name="email"]')->sendKeys('demo@example.com');

sendKeys is the WebDriver way: simulates keystrokes one at a time. The Symfony form helpers are higher-level, pick whichever fits your interaction shape.

When Panther beats Playwright-via-subprocess

You could call Playwright Python from PHP via exec() or Symfony\Process. Panther is better when:

1. Your project is already Symfony. Panther integrates with Symfony's test framework (PantherTestCase), the dependency injection container, console commands, and the dotenv config. Lesson 2.13 builds a Symfony Console scraper that's idiomatic in a way subprocess-Playwright never could be.

2. You need transactional control. Panther lives in your PHP process, sharing config, services, and exceptions. A subprocess Python script is a black box, communicate over stdout, hope for the best.

When Playwright (Python or Node) is still better:

1. The team's primary language is Python/JS and PHP is one consumer.
2. You need Playwright-specific features (browser context proxies, request interception, codegen).
3. You're scraping at high volume where every millisecond counts, Panther's WebDriver protocol has more round-trips than CDP.

Cleanup

try {
  $client = Client::createChromeClient();
  // ... scraping ...
} finally {
  $client->quit();
}

ChromeDriver and Chromium are subprocesses. Forgetting quit() leaves them running, sometimes outliving your PHP process. Always wrap in try/finally.

A note on environment

Headless Chromium needs:

• A modern Linux/macOS/Windows kernel.
• For CI/Docker: ensure the standard fontconfig + libnss libraries are installed, otherwise Chromium errors on startup.
• For dockerised PHP: use a base image that's known to work (webdevops/php-nginx series ships everything; vanilla php:fpm does not).

The first run downloads ~150 MB of driver/browser. Subsequent runs are instant.

Hands-on lab

Open /challenges/dynamic/spa-pure. Write a Panther scraper that: (1) navigates to the page, (2) waits for the product grid to render, (3) extracts the name and price of every product card. Then add a click on "Load more" if the lab supports it, and re-extract. You should have a working browser-driven PHP scraper in under thirty lines.