Skip to content

Bring your Playwright script

Your Playwright script works until a site starts blocking it. Pydoll runs the same script over raw CDP, with no driver process and no patched browser, so there is nothing for a detector to find in the transport and nothing for you to rewrite. This page takes an existing Playwright script and gets it running on Pydoll in five minutes, then shows where Pydoll goes further than Playwright when you need it.

You will learn

Swap the import

Replace playwright.sync_api with pydoll.playwright.sync_api, or playwright.async_api with pydoll.playwright.async_api. Everything else stays. You don't install the playwright package or its bundled browsers; Pydoll drives the Chrome or Edge already on your machine.

from pydoll.playwright.sync_api import sync_playwright

with sync_playwright() as p:
    browser = p.chromium.launch()
    page = browser.new_page()
    page.goto('https://quotes.toscrape.com/login')

    page.get_by_label('Username').fill('john')
    page.locator('#password').fill('SecretPass123')
    page.get_by_role('button', name='Login').click()
    page.wait_for_url('**/')

    print(page.get_by_role('link', name='Logout').is_visible())
    browser.close()

import asyncio

from pydoll.playwright.async_api import async_playwright


async def main():
    async with async_playwright() as p:
        browser = await p.chromium.launch()
        page = await browser.new_page()
        await page.goto('https://quotes.toscrape.com/login')

        await page.get_by_label('Username').fill('john')
        await page.locator('#password').fill('SecretPass123')
        await page.get_by_role('button', name='Login').click()
        await page.wait_for_url('**/')

        print(await page.get_by_role('link', name='Logout').is_visible())
        await browser.close()

asyncio.run(main())

Run it the way you always did. Locators stay lazy and strict, get_by_role resolves the same elements, actions auto-wait for the element to be attached, visible, stable and enabled, and a failed wait raises the same TimeoutError with the same call log.

chromium.launch() takes the options you already pass: headless, args, executable_path, proxy, downloads_path. launch_persistent_context(user_data_dir) and connect_over_cdp(url) work too. firefox and webkit raise an error, because Pydoll drives Chromium-based browsers only.

Drop-in for the automation surface

The layer covers what a script uses: Playwright, Browser, BrowserContext, Page, Frame, Locator, ElementHandle, Keyboard, Mouse, routes, dialogs and downloads. It does not cover the test runner: expect() assertions, fixtures, the pytest plugin, tracing and video are not implemented. Playwright API lists every class with what is full, partial or missing.

What changed under your script

Playwright talks to the browser through a driver process written in Node. Every call crosses into that process, which then speaks CDP to Chrome. Stealth forks such as Patchright patch that driver after the fact, removing the calls that give automation away.

Pydoll has no driver. Your page.goto becomes CDP commands sent straight from Python, over the same connection Pydoll's own API uses. That removes the transport tells at the source instead of patching them:

  • The Runtime domain, whose activation is the best-known automation tell, is never enabled for your script. Selectors and actionability checks run in an isolated world of each frame, evaluated once, and your page.evaluate code runs in the main world without going through eval.
  • No --enable-automation, no navigator.webdriver, no console hooks, no init scripts injected on every navigation.
  • Evaluations carry no synthetic user gesture, popups open only from a real user action, and a user_agent override ships coherent Client Hints, as in a browser someone is using.

The rest of the browser is a real Chrome: its TLS, HTTP/2 and rendering fingerprints are the ones Chrome has, because it is Chrome.

Reach Pydoll from a Playwright object

Every Playwright object exposes the Pydoll object underneath, so features Playwright has no name for are one attribute away. You don't rewrite the script to use them.

Playwright object Pydoll object underneath
browser.chrome Chrome
page.tab Tab
locator.element_handle().web_element WebElement

Three things a blocked script usually needs next, each one line:

from pydoll.sync import ExtractionModel, Field
from pydoll.playwright.sync_api import sync_playwright

from examples.fingerprints import FINGERPRINTS


class Quote(ExtractionModel):
    text: str = Field(selector='.text')
    author: str = Field(selector='.author')


with sync_playwright() as p:
    browser = p.chromium.launch()
    page = browser.new_page()

    page.tab.apply_fingerprint(FINGERPRINTS['macos_m3_new_york'])
    with page.tab.expect_cloudflare_turnstile():
        page.goto('https://quotes.toscrape.com')

    quotes = page.tab.extract_all(Quote, scope='.quote')
    submit = page.get_by_role('link', name='Login').element_handle().web_element
    submit.click(humanize=True)

from pydoll import ExtractionModel, Field
from pydoll.playwright.async_api import async_playwright

from examples.fingerprints import FINGERPRINTS


class Quote(ExtractionModel):
    text: str = Field(selector='.text')
    author: str = Field(selector='.author')


async with async_playwright() as p:
    browser = await p.chromium.launch()
    page = await browser.new_page()

    await page.tab.apply_fingerprint(FINGERPRINTS['macos_m3_new_york'])
    async with page.tab.expect_cloudflare_turnstile():
        await page.goto('https://quotes.toscrape.com')

    quotes = await page.tab.extract_all(Quote, scope='.quote')
    submit = (await page.get_by_role('link', name='Login').element_handle()).web_element
    await submit.click(humanize=True)

apply_fingerprint sets a coherent identity across every layer (Fingerprint injection), expect_cloudflare_turnstile clicks the Turnstile widget if the navigation inside its block meets one (Captcha bypass), extract_all returns typed, validated objects (Structured extraction), and humanize=True moves the mouse along a curved path with human timing (Human-like interactions). Raw CDP is one call away as well: page.tab.execute_command(...).

What this does not change

What detection still sees

Pydoll removes the automation tells a library adds; it does not make a bot a person. A headless browser is still recognizable by its rendering and its missing media devices, so run headful or apply a fingerprint profile that covers the headless signals. Your IP reputation counts as much as the browser; a datacenter IP fails challenges a residential one passes. Signals baked into the Chrome binary are not touched. And behavior matters: instant clicks and perfectly regular typing look like what they are, which is what humanize=True is for.

Staying undetected walks through each layer with the minimum setup for it.

What's next