Pydoll is a stealth-first Python library that automates Chromium-based browsers directly over the DevTools Protocol, without a WebDriver, for developers who scrape, test, or automate sites that block ordinary bots.

What it is

Pydoll is a browser automation library for Python that lives in the web scraping and end-to-end testing ecosystem. It drives Chrome directly over the Chrome DevTools Protocol (CDP) using a WebSocket connection, so there is no driver binary in the loop. It sits alongside tools such as Playwright and Puppeteer as a Chromium automation option, but it is written for Python from the ground up on top of asyncio, with type checking through mypy.

The concrete problem it solves is the wall that appears once automation leaves a developer machine. A scraper that runs fine locally tends to hit captchas and Cloudflare challenges when it runs for real, because standard automation leaks signals: the WebDriver binary, the navigator.webdriver flag, and inconsistent browser fingerprints. Pydoll removes the driver binary entirely and also tries to make the browser present a coherent identity, so ordinary bot protection has less to detect. Its own documentation is explicit that results on behavioral challenges depend on the browser and the IP reputation behind it, so it is a mitigation rather than a guarantee.

Key capabilities

• Fingerprint injection through tab.apply_fingerprint(), aligning User-Agent, Client Hints, navigator, WebGL, canvas, screen, fonts, timezone, and locale. The overrides survive toString and prototype introspection and propagate into Web Workers, so lie-detection checks such as CreepJS's do not flag them.
• Humanized interactions: mouse movement along Bezier curves, realistic typing, and scroll physics.
• Zero WebDrivers — a direct CDP connection over WebSocket, with no driver binary and no navigator.webdriver flag.
• An async, fully typed API built on asyncio and checked with mypy, giving IDE autocompletion and static error checking.
• Network control for intercepting requests to block ads and trackers, monitoring traffic for API discovery, and making authenticated HTTP requests that inherit the browser session.
• Full support for shadow roots, including closed ones, and for cross-origin iframes, queried and driven through the same API.
• Structured extraction: define a Pydantic model, call tab.extract(), and receive typed, validated data instead of querying element by element.

Who uses it and how

• Scraping teams whose pipelines break on captchas and Cloudflare challenges in production but work locally, and who need the browser to look consistent rather than merely headless.
• QA and test engineers writing end-to-end tests against Chromium, using the async API inside existing Python test suites.
• Crawler and anti-detection work where the browser must pass behavioral challenges such as Cloudflare Turnstile or reCAPTCHA v3, in combination with a reputable IP.
• Projects that need to reach inside shadow DOM, including closed shadow roots, and cross-origin iframes without switching tooling.
• Data extraction pipelines that want validated output from a Pydantic model rather than manual parsing of individual elements.

Getting started

The README points to the Documentation and Getting Started pages at https://pydoll.tech/, and the library requires Python 3.10 or newer. No Docker image or hosted option is described; it is installed and used as a Python library.

How it compares

Among the similar tools named in its own topics, Pydoll is grouped with Playwright and Puppeteer as Chromium automation, but it differs in mechanism: it uses a direct CDP connection over WebSocket rather than a WebDriver, and it ships fingerprint injection and humanized input as first-class features. Developers already committed to a JavaScript test runner have little reason to move, while Python teams that have been fighting detection have a more targeted option.

When to use it — and when not to

Choose Pydoll when the target is Chromium and the pain is detection, fingerprints, or driver version mismatches. Be aware that it is maintained by a single person who states plainly that releases and issue replies may be slower than usual, and that Firefox support is a stated goal tied to reaching 10k stars rather than a shipped feature. It is also the wrong pick for anyone who cannot operate their own browser environment and reputable IP, since passing behavioral challenges depends on both.

pip install pydoll-python