Playwright API

pydoll.playwright is a Playwright-compatible API for the browser automation surface: the same classes, methods, selectors and events your Playwright Python script already uses, executed over Pydoll's CDP connection with no driver process. This page is the contract: what is implemented in full, what works with a named difference, and what is missing, so you know before you switch. For the five-minute version, start with Bring your Playwright script.

The rule that governs the list: if a browser automation script works on Playwright and not here, it is a bug, and an issue with the script is welcome. Test-runner features are not bugs; they are out of scope.

Import paths

Playwright	Pydoll
`from playwright.sync_api import sync_playwright`	`from pydoll.playwright.sync_api import sync_playwright`
`from playwright.async_api import async_playwright`	`from pydoll.playwright.async_api import async_playwright`
`from playwright.sync_api import Page, Locator, ...`	`from pydoll.playwright.sync_api import Page, Locator, ...`
`playwright.sync_api.TimeoutError`, `Error`	`pydoll.playwright.sync_api.TimeoutError`, `Error`

The sync module is generated from the async one, the way Playwright's own is, so the two never differ in methods or arguments.

Compatibility matrix

Full means the same signature and behavior. Partial means it works and the difference is named next to it. Missing means not implemented, with the Pydoll way to do it when there is one.

Playwright and BrowserType

Member	Status	Notes
`chromium.launch()`	Full	`headless`, `args`, `executable_path`, `proxy`, `downloads_path` map onto ChromiumOptions
`chromium.launch_persistent_context()`	Full
`chromium.connect_over_cdp()`	Full
`selectors.set_test_id_attribute()`	Full
`devices`	Full
`firefox`, `webkit`	Missing	Pydoll drives Chromium-based browsers only; raises `Error`
`chromium.connect()`	Missing	The Playwright server protocol has no equivalent; use `connect_over_cdp`
`request` (`APIRequestContext`)	Missing	Use browser-context HTTP requests through `page.tab.request`

Accepted and ignored on launch(): slow_mo, devtools, env, ignore_default_args, channel, chromium_sandbox, handle_sigint and friends.

Browser

Member	Status	Notes
`new_context()`, `new_page()`, `contexts`, `version`, `close()`	Full
`disconnected` event	Partial	Fires when you call `close()`; it does not fire when the browser process dies on its own
`new_browser_cdp_session()`	Missing	Use `browser.chrome.execute_command(...)`
`start_tracing()`, `stop_tracing()`	Missing	Out of scope

BrowserContext

Member	Status	Notes
`new_page()`, `pages`, `close()`	Full
`cookies()`, `add_cookies()`, `clear_cookies()`	Full
`storage_state()`	Full	Read and write
`grant_permissions()`, `clear_permissions()`, `set_geolocation()`	Full
`set_extra_http_headers()`, `set_offline()`	Full
`add_init_script()`, `expose_function()`, `expose_binding()`	Full
`route()`, `unroute()`, `route_from_har()`	Partial	`route` and `unroute` are full; `route_from_har` is missing
`expect_page()`, `expect_event()`, `wait_for_event()`	Full
`set_default_timeout()`, `set_default_navigation_timeout()`	Full
Options: `viewport`, `user_agent`, `locale`, `timezone_id`, `extra_http_headers`, `storage_state`, `permissions`, `offline`, `base_url`, `http_credentials`, `ignore_https_errors`, `java_script_enabled`, `bypass_csp`, `color_scheme`, `device_scale_factor`, `is_mobile`, `has_touch`	Full	`http_credentials` answer the `401` challenge through the Fetch domain rather than sending a fixed header
Options: `record_video_dir`, `record_har_path`, `client_certificates`, `service_workers`, `strict_selectors`, `accept_downloads=False`	Partial	Accepted and ignored
`new_cdp_session()`	Missing	Use `page.tab.execute_command(...)`
`request` (`APIRequestContext`)	Missing	Use `page.tab.request`
`tracing`, `clock`	Missing	Out of scope

Page

Member	Status	Notes
`goto()`, `reload()`, `go_back()`, `go_forward()`	Full	`wait_until='commit'` resolves when the new document commits
`wait_for_load_state()`, `wait_for_url()`, `expect_navigation()`	Full
`evaluate()`, `evaluate_handle()`	Partial	Runs in the main world without `eval`; `Date` comes back as an ISO string, `undefined` and `null` both as `None`
`query_selector()`, `query_selector_all()`, `wait_for_selector()`, `wait_for_function()`, `wait_for_timeout()`	Full
`content()`, `set_content()`, `title()`, `url`	Full
`frames`, `main_frame`, `frame()`, `frame_locator()`	Partial	`frames` fills from frame events; a child frame reports its parent's document until it resolves, shortly after attach
`add_init_script()`, `add_script_tag()`, `add_style_tag()`	Full
`set_viewport_size()`, `viewport_size`, `emulate_media()`, `set_extra_http_headers()`	Full
`screenshot()`, `pdf()`	Partial	`mask`, `animations`, `caret`, `scale` and `style` options of `screenshot` are accepted and ignored
`route()`, `unroute()`, `expose_function()`, `expose_binding()`	Full
`keyboard`, `mouse`, `touchscreen`	Full	Playwright key names with the US layout: `Enter`, `Control+A`, `KeyA`, `Digit1`
Selector shortcuts: `click()`, `fill()`, `type()`, `press()`, `check()`, `select_option()`, `text_content()`, `inner_text()`, `inner_html()`, `get_attribute()`, `is_visible()`, ...	Full
`get_by_role()`, `get_by_text()`, `get_by_label()`, `get_by_placeholder()`, `get_by_alt_text()`, `get_by_title()`, `get_by_test_id()`	Full	ARIA roles and accessible names computed with a port of Playwright's algorithm
Events: `load`, `domcontentloaded`, `framenavigated`, `request`, `response`, `requestfinished`, `requestfailed`, `dialog`, `console`, `pageerror`, `download`, `popup`, `filechooser`, `close`, `crash`	Full	With the matching `expect_*` context managers; `console` and `pageerror` are the only listeners that enable the `Runtime` domain
Events: `websocket`, `worker`	Missing
`wait_for_request()`, `wait_for_response()`	Missing	Use `expect_request()` and `expect_response()`
`pause()`, `add_locator_handler()`, `aria_snapshot()`, `clock`	Missing	Out of scope
`request` (`APIRequestContext`)	Missing	Use `page.tab.request`

Frame, FrameLocator, Locator, ElementHandle, JSHandle

Member	Status	Notes
The full method set, including `filter()`, `and_()`, `or_()`, `nth()`, `first`, `last`, `count()`, `all()`, `drag_to()`, `select_option()`, `set_input_files()`, `screenshot()`, `evaluate_all()`, `bounding_box()`, `scroll_into_view_if_needed()`, `dispatch_event()`	Full	Locators are lazy and strict; actions follow Playwright's actionability checks and produce the same call log on timeout
`set_input_files()`	Full	Paths and `FilePayload`
`highlight()`, `aria_snapshot()`	Missing

Selectors

Syntax	Status	Notes
CSS, XPath, `text=`, `xpath=`, `css=`, `id=`, `data-testid=`, `nth=`, `visible=`, chaining with `>>`, `internal:` engines behind `get_by_*`	Full
Playwright CSS pseudo-classes `:has-text()`, `:visible`, `:nth-match()`, `:text()`	Missing	Use `get_by_text()`, `filter(has_text=...)`, `locator(...).nth()`
Layout selectors `:left-of()`, `:right-of()`, `:above()`, `:below()`, `:near()`	Missing
`react=`, `vue=` engines, custom selector engines	Missing

Request, Response, Route, Dialog, Download, ConsoleMessage, FileChooser

Full. Routing runs on the Fetch domain that request interception uses. route.fetch() lets the request reach the network and pauses again at the response stage, so a handler can rewrite a body with route.fulfill(response=..., body=...).

Not in scope

expect() assertions, the pytest-playwright plugin and its fixtures, tracing, video recording, HAR replay, Clock, Tracing, Worker, WebSocket, Android, Electron. A script that needs the test runner keeps using Playwright for the runner; the browser automation inside a test can still be Pydoll through connect_over_cdp.

Behaviors that differ on purpose

These come from Pydoll's stealth defaults and are kept under the Playwright surface:

Evaluations run without a synthetic user gesture, so navigator.userActivation stays false until a real click. Set user_gesture_on_evaluate=True on the context if a script needs window.open from evaluate.
Popups are blocked unless a user action opens them, as in a normal Chrome.
Actions do not wait for a navigation they start, and no_wait_after is accepted and ignored. When a click submits a form or follows a link, wait with page.wait_for_url() or page.expect_navigation() before reading the new page.
A user_agent override ships coherent Client Hints and follows Chrome's reduced UA format; the string you pass is normalized the way Pydoll does for its own options. locale sets a Chrome-shaped Accept-Language.
http_credentials only answer a challenge from their own origin.
A cross-origin iframe that Chrome renders out of process cannot be reached through frame_locator() or a selector chain that enters the frame: the selector engine lives in the parent's session while the query would have to run in the child's. Use element_handle.content_frame() and the frame's own methods for those frames.
Context-level init scripts, bindings and emulation reach a popup only after it is adopted, so they miss the popup's very first document.
Viewport emulation keeps screen at least as large as the viewport, so the two never contradict each other.
The selector and actionability engine runs in an isolated world of each frame, evaluated once and reused, so a page that wraps DOM prototypes, requestAnimationFrame or window.eval never sees it. Your own page.evaluate code runs in the main world, where the page's globals live, and its source is embedded in the call rather than passed through eval.

Reaching Pydoll

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

SyncAsync

with page.tab.expect_cloudflare_turnstile():
    page.goto('https://example.com/protected')

button = page.get_by_role('button', name='Continue').element_handle().web_element
button.click(humanize=True)

async with page.tab.expect_cloudflare_turnstile():
    await page.goto('https://example.com/protected')

button = (await page.get_by_role('button', name='Continue').element_handle()).web_element
await button.click(humanize=True)

What's next

Bring your Playwright script: the walkthrough, with the honest note on what detection still sees.
Request interception: the Fetch domain that routes are built on.
Migrating from Selenium and Playwright: the same moves in Pydoll's own API.