🪄 Patches for undetectable browser automation

This repo contains patches to enhance popular web automation libraries. Specifically, it targets the puppeteer and playwright packages.

Some aspects of automation libraries or browser behavior cannot be adjusted through settings or command-line switches. Therefore, we fix these issues by patching the library's source code. While this approach is fragile and may break as the libraries' source code changes over time, the goal is to maintain this repo with community help to keep the patches up to date.

Do I really need any patches?

Out of the box Puppeteer and Playwright come with some significant leaks that are easy to detect. It doesn't matter how good your proxies, fingeprints, and behaviour scripts, if you don't have it patched, you're just a big red flag for any major website.

🕵️ You can easily test your automation setup for major modern detections with rebrowser-bot-detector (sources and details)

Before the patches 👎	After the patches 👍

Is there an easy drop-in replacement?

If you don't want to mess with the patches and all possible errors, there is a drop-in solution for you. These packages have simply applied rebrowser-patches on top of the original code, nothing more.

Puppeteer: rebrowser-puppeteer (source mirror) and rebrowser-puppeteer-core (source mirror)

Playwright (Node.js): rebrowser-playwright (source mirror) and rebrowser-playwright-core (source mirror)

Playwright (Python): rebrowser-playwright (source mirror)

The easiest way to start using them is to install them under the original package names (an npm alias). This way, you don't need to change any source code of your automation. Install only the packages your project already uses:

# Puppeteer
npm install puppeteer@npm:rebrowser-puppeteer puppeteer-core@npm:rebrowser-puppeteer-core

# Playwright
npm install playwright@npm:rebrowser-playwright playwright-core@npm:rebrowser-playwright-core

npm installs the latest version and saves it in package.json as "puppeteer": "npm:rebrowser-puppeteer@^…". With yarn or pnpm, run yarn add or pnpm add with the same arguments.

Another way is to use the new package names directly: install rebrowser-puppeteer and rebrowser-puppeteer-core (or rebrowser-playwright and rebrowser-playwright-core) instead of the original packages, then replace puppeteer with rebrowser-puppeteer (and so on) in your imports.

For Python, run pip install rebrowser-playwright and import from rebrowser_playwright instead of playwright, e.g. from rebrowser_playwright.sync_api import sync_playwright.

The packages follow the version numbers of the original libraries. A fix for a version that is already out ships as X.Y.100, X.Y.101 and so on, so always install the latest version instead of matching the version of the original package.

🚀 That's it! Just visit the rebrowser-bot-detector page and test your patched browser.

Notes:

• The drop-ins patch the library code only, not the browser. Install a browser the same way as with the original package: rebrowser-puppeteer downloads Chrome on install, rebrowser-playwright needs npx playwright install chromium (or use executablePath / channel: 'chrome' with a Chrome you already have).
• Since 1.1.1 (drop-ins 25.14.100 / 1.64.100), rebrowser-puppeteer and rebrowser-playwright depend on rebrowser-puppeteer-core / rebrowser-playwright-core under their real names. Earlier versions used an npm alias, and when the original puppeteer / playwright of the same version was installed in the same project, npm quietly gave the drop-in the unpatched core (#74).

Our goal is to maintain and support these drop-in replacement packages with the latest versions, but we mainly focus on fresh versions, so if you're still using puppeteer 13.3.7 from the early 90s, it might be a good time to upgrade. There's a high chance that it won't really break anything as the API is quite stable over time.

Available patches

Fix Runtime.Enable leak

Popular automation libraries rely on the CDP command Runtime.Enable, which allows receiving events from the Runtime. domain. This is crucial for managing execution contexts used to evaluate JavaScript on pages, a key feature for any automation process.

However, there's a technique that detects the usage of this command, revealing that the browser is controlled by automation software like Puppeteer or Playwright. This technique is used by all major anti-bot software such as Cloudflare, DataDome, and others.

We've prepared a full article about our investigation on this leak, which you can read in our blog.

For more details on this technique, read DataDome's blog post: How New Headless Chrome & the CDP Signal Are Impacting Bot Detection. In brief, it's a few lines of JavaScript on the page that are automatically called if Runtime.Enable was used.

Our fix disables the automatic Runtime.Enable command on every frame. Instead, we manually create contexts with unknown IDs when a frame is created. Then, when code needs to be executed, there are multiple ways to get the context ID.

1. Look up the main world's context ID from the DevTools side (mode addBinding, the default).

The patch reads the frame's document in an isolated world, resolves that node again in the frame's main world and takes the context ID from the result. Nothing runs in the page, and nothing is added to it. Only if that lookup fails does it fall back to the original technique: create a binding in the main world, call it and save its context ID. The fallback cleans up after itself, but a page could briefly see a global with a random name.

🟢 Pros: The ultimate approach that keeps access to the main world and works with web workers and iframes. You don't need to change any of your existing codebase.

🔴 Cons: None are discovered so far.

2. Create a new isolated context via Page.createIsolatedWorld and save its ID.

🟢 Pros: All your code will be executed in a separate isolated world, preventing page scripts from detecting your changes via MutationObserver and other techniques.

🔴 Cons: You won't be able to access main context variables and code. While this is necessary for some use cases, the isolated context generally works fine for most scenarios. Also, web workers don't allow creating new worlds, so you can't execute your code inside a worker. This is a niche use case but may matter in some situations. There is a workaround for this issue, please read How to Access Main Context Objects from Isolated Context in Puppeteer & Playwright.

3. Call Runtime.Enable and then immediately call Runtime.Disable.

This triggers Runtime.executionContextCreated events, allowing us to catch the proper context ID.

🟢 Pros: You will have full access to the main context.

🔴 Cons: There's a slight chance that during this short timeframe, the page will call code that leads to the leak. The risk is low, as detection code is usually called during specific actions like CAPTCHA pages or login/registration forms, typically right after the page loads. Your business logic is usually called a bit later.

🎉 Our tests show that all these approaches are currently undetectable by Cloudflare or DataDome.

Note: you can change settings for this patch on the fly using an environment variable. This allows you to easily switch between patched and non-patched versions based on your business logic.

• REBROWSER_PATCHES_RUNTIME_FIX_MODE=addBinding — addBinding technique (default). Keeps main-world access and works with iframes and web workers.
• REBROWSER_PATCHES_RUNTIME_FIX_MODE=alwaysIsolated — always run all scripts in an isolated context. Page scripts can't observe your execution, but you lose main-world access, page.exposeFunction() and (Puppeteer) web workers.
• REBROWSER_PATCHES_RUNTIME_FIX_MODE=enableDisable — use the Enable/Disable technique. Not recommended: it still calls Runtime.enable for a brief window, so it is not fully leak-free. Only use it if addBinding fails on a specific page.
• REBROWSER_PATCHES_RUNTIME_FIX_MODE=0 — completely disable fix for this leak
• REBROWSER_PATCHES_DEBUG=1 — enable some debugging messages

Remember, you can set these variables in different ways, for example, in code:

process.env.REBROWSER_PATCHES_RUNTIME_FIX_MODE = "alwaysIsolated"

or in command line:

REBROWSER_PATCHES_RUNTIME_FIX_MODE=alwaysIsolated node app.js

Keep library names out of stack traces

By default, Puppeteer adds //# sourceURL=pptr:... to every script in page.evaluate(), and Playwright runs page.evaluate() through a helper class named UtilityScript. A page that takes new Error().stack while your code runs (for example from a getter your code touches) sees those names and knows it's being automated.

The Puppeteer patch adds no sourceURL at all, so frames show `` like any other injected script. The Playwright patch renames the helper to a short name picked at random per process and shifts its line numbers, so there's no fixed name or position to look for. You can adjust it via environment variable:

# Puppeteer only: use a sourceURL of your choice. Prefer a full URL: a bare file name such as
# jquery.min.js shows up in stack traces as a script with no address, which detectors can flag
REBROWSER_PATCHES_SOURCE_URL=https://example.com/js/app.js
# use 0 to completely disable this patch (Puppeteer and Playwright)
REBROWSER_PATCHES_SOURCE_URL=0

Hide navigator.webdriver

Chrome launched by Puppeteer or Playwright reports navigator.webdriver = true — an obvious automation flag that many anti-bot systems check. This patch adds --disable-blink-features=AutomationControlled to the default launch arguments so navigator.webdriver stays false. Any --disable-blink-features=... you pass yourself is merged in, not overwritten.

Puppeteer also passes --enable-automation, which shows the "Chrome is being controlled by automated test software" bar. The Puppeteer patch drops it (Playwright stopped passing it in 1.64).

Both libraries also launch Chrome with switches that a page can notice, so the patches drop them too:

• --disable-popup-blocking: a page could open popups without a click. Now Chrome blocks popups the way it does for a real user: page.click() and page.evaluate() count as a click, so the first popup they open still opens, but a second one from the same click, one opened after a few seconds or one opened by the page on load is blocked. To bring the old behavior back, pass --disable-popup-blocking in args.
• --disable-ipc-flooding-protection: Chrome normally ignores history updates past 200 in 10 seconds; with the switch, a page sees all of them applied. Only pages that change the URL more than ~20 times a second notice the difference. Pass the switch in args to keep it.
• Playwright only, ThirdPartyStoragePartitioning in --disable-features: an iframe from another site that embeds a frame from the page's own site could read the page's localStorage. Now third-party storage is partitioned as in a normal Chrome. Playwright disabled partitioning so that localStorage seeded with storageState also shows up when that site is embedded in another site's page (microsoft/playwright#32230); if you rely on that, set REBROWSER_PATCHES_WEBDRIVER_FIX=0 and pass --disable-blink-features=AutomationControlled in args yourself, which keeps navigator.webdriver hidden. Passing your own --disable-features=ThirdPartyStoragePartitioning instead would replace Playwright's whole --disable-features list (Chrome keeps only the last one).

This applies to Chrome launched with puppeteer.launch() and Playwright's chromium.launch() / launchPersistentContext(); Playwright's Android and Electron launchers are unchanged.

# use 0 to disable this patch and keep the library's default behavior
REBROWSER_PATCHES_WEBDRIVER_FIX=0

Note: this only affects browsers launched by the library. If you connect to an existing browser, pass the switch yourself.

Method to access browser CDP connection

Sometimes, it could be very useful to access a CDP session at a browser level. For example, when you want to implement some custom CDP command. There is a method page._client() that returns CDP session for the current page instance, but there is no such method for browser instance. This patch adds a new method _connection() to Browser class, so you can use it like this:

browser._connection().on('Rebrowser.addRunEvent', (params) => { ... })

Note: it's not detectable by external website scripts, it's just for your convenience.

Change default utility world name

The default utility world name is '__puppeteer_utility_world__' + packageVersion. Sometimes you might want to change it to something else. This patch changes it to util and allows you to customize it via env variable:

REBROWSER_PATCHES_UTILITY_WORLD_NAME=customUtilityWorld
# use 0 to completely disable this patch
REBROWSER_PATCHES_UTILITY_WORLD_NAME=0

This env variable cannot be changed on the fly, you have to set it before running your script because it's used at the moment when the module is getting imported.

Before patch 👎	After patch 👍

Note: it's not detectable by external website scripts, but Google might use this information in their proprietary Chrome; we never know.

Usage

This package is designed to be run against an installed library. Install the library, then call the patcher, and it's ready to go.

In the root folder of your project, run:

npx rebrowser-patches@latest patch --packageName puppeteer-core

You can easily revert all changes with this command:

npx rebrowser-patches@latest unpatch --packageName puppeteer-core

You can also patch a package by providing the full path to its folder, for example:

npx rebrowser-patches@latest patch --packagePath /web/app/node_modules/puppeteer-core-custom

The patcher ships a separate patch set per upstream version (under patches///). It picks the patch that matches your installed version exactly; if there isn't one, it tries the newest patch for the same major.minor, then the newest overall, and warns that the version wasn't tested. If that closest patch doesn't apply cleanly it stops without touching your files and suggests the matching drop-in package. For the smoothest experience, keep your library on a version this repo lists as tested, or use the drop-in packages below.

You can see all command-line options by running npx rebrowser-patches@latest --help, but currently, there's just one patch for one library, so you don't need to configure anything.

⚠️ Be aware that after running npm install or yarn install in your project folder, it might override all the changes from the patches. You'll need to run the patcher again to keep the patches in place.

How to update the patches?

If you already have your package patched and want to update to the latest version of rebrowser-patches, the easiest way would be to delete node_modules/puppeteer-core, then run npm install or yarn install --check-files, and then run npx rebrowser-patches@latest patch.

How to patch Java/Python/.NET versions of Playwright?

All these versions are just wrappers around Node.js version of Playwright. You need to find driver folder inside your Playwright package and run this patch with --packagePath=$yourDriverFolder/$yourPlatform/package.

Puppeteer support

✅ Latest fully tested version: 25.14.0 (released 2026-10-14), drop-in packages 25.14.101

Playwright support

Playwright patches include:

• Runtime.enable leak: addBinding and alwaysIsolatedmodes.
• Ability to change utility world name via REBROWSER_PATCHES_UTILITY_WORLD_NAME env variable.
• More patches are coming, star and follow the repo.

Important: page.pause() method doesn't work with the enabled fix, it needs more investigation. You can just disable the fix completely while debugging using REBROWSER_PATCHES_RUNTIME_FIX_MODE=0 env variable.

The patches target Chromium. On Firefox and WebKit the Chromium-specific context handling is skipped, so one install can drive all three browsers (before 1.64.1 it crashed with _sessions is undefined). Only small shared changes, such as the renamed UtilityScript helper, apply there too.

✅ Latest fully tested version: 1.64.1 (released 2026-10-14), drop-in packages 1.64.101 (Python: 1.64.101, built from playwright-python 1.64.0)

How to use rebrowser-puppeteer with puppeteer-extra?

Use addExtra method, here is the example:

// before
import puppeteer from 'puppeteer-extra'

// after
import { addExtra } from 'puppeteer-extra'
import rebrowserPuppeteer from 'rebrowser-puppeteer-core'
const puppeteer = addExtra(rebrowserPuppeteer)

Be careful with puppeteer-extra-plugin-stealth: its evasions override browser internals from JavaScript, and that is detectable on its own. On rebrowser-bot-detector rebrowser-puppeteer with no plugins comes out clean, while adding the stealth plugin turns nativeTampering and workerConsistency red.

Releases and versions

• Every package is built from this repo and published by GitHub Actions with npm and PyPI trusted publishing, so npm shows provenance for each version. Release notes are in GitHub releases.
• The drop-in packages follow the upstream version. A version like 25.14.101 is a re-release of upstream 25.14.0 with our fixes (patch numbers from 100 up), so it never collides with an upstream release.
• In each source mirror (rebrowser-puppeteer, rebrowser-playwright, rebrowser-playwright-python and the -core ones), every version has its own branch with two commits: the original package, then the patched one. The second commit's diff is exactly what changed.

Follow the project

We're currently developing more patches to improve web automation transparency, which will be released in this repo soon. Please support the project by clicking ⭐️ star or watch button.

💭 If you have any ideas, thoughts, or questions, feel free to reach out to our team by email or use the issues section.

The fix doesn't help, I'm still getting blocked 🤯

⚠️ It's important to know that this fix alone won't make your browser bulletproof and undetectable. You need to address many other aspects such as proxies, proper user-agent and fingerprints (canvas, WebGL), and more.

What the patches cover: the Runtime.enable leak, library names in stack traces and world names, navigator.webdriver and the automation infobar. What they don't touch: the HeadlessChrome token in the headless user agent, fingerprints (canvas, WebGL, fonts, screen), your IP reputation and how your script behaves (timing, mouse movement). CAPTCHA and bot-management products such as Cloudflare Turnstile, hCaptcha or Google search score all of these, so the patches alone won't get you past them.

Always keep in mind: the less you manipulate browser internals via JS injections, the better. There are ways to detect that internal objects such as console, navigator, and others were affected by Proxy objects or Object.defineProperty. It's tricky, but it's always a cat-and-mouse game.

If you've tried everything and still face issues, try asking a question in the issues section or consider using cloud solutions from Rebrowser.

What is Rebrowser?

This package is sponsored and maintained by Rebrowser. We allow you to scale your browser automation and web scraping in the cloud with hundreds of unique fingerprints.

Our cloud browsers have great success rates and come with nice features such as notifications if your library uses Runtime.Enable during execution or has other red flags that could be improved. Create an account today to get invited to test our bleeding-edge platform and take your automation business to the next level.

Patch command on Windows

When you try to run this patcher on a Windows machine, you will probably encounter an error because the patch command is not found. To fix this, you need to install Git, which includes patch.exe. After you have installed it, you need to add it to your PATH:

set PATH=%PATH%;C:\Program Files\Git\usr\bin\

You can check that patch.exe is installed correctly by using next command:

patch -v

Special thanks

zfcsoftware/puppeteer-real-browser - general ideas and contribution to the automation community

Kaliiiiiiiiii-Vinyzu/patchright - set of patches to fix Playwright leaks

kaliiiiiiiiii/brotector - some modern tests, algorithm to distinguish CDP vs devtools

prescience-data/harden-puppeteer - one of the pioneers of the execution in an isolated world

puppeteer-extra-plugin-stealth - where it all started, big props to all the contributors and the community 🙏 berstend and co are the goats

Disclaimer

No responsibility is accepted for the use of this software. This software is intended for educational and informational purposes only. Users should use this software at their own risk. The developers of the software cannot be held liable for any damages that may result from the use of this software. This software is not intended to bypass any security measures, including but not limited to CAPTCHAs, anti-bot systems, or other protective mechanisms employed by websites. The software must not be used for malicious purposes. By using this software, you agree to this disclaimer and acknowledge that you are using the software responsibly and in compliance with all applicable laws and regulations.