antibrow 0.1.0

The antidetect browser your AI agent can drive - kernel-level fingerprint spoofing with the Playwright API

antibrow

The antidetect browser your AI agent can drive.

Kernel-level fingerprint spoofing · unlimited local profiles, free · the Playwright API you already write

from antibrow import launch

browser = launch()
page = browser.new_page()
page.goto("https://abrahamjuliot.github.io/creepjs/")

That is a real Chromium — with a real device's fingerprint, its own persistent profile, and no playwright install step — driven by the exact Playwright API you already know.

---

Contents

• Why antibrow
• Install
• Quick start
• API reference
• Profiles and fingerprints
• Proxies
• Framework integrations
  • Playwright · Puppeteer / Node · browser-use · crawl4ai · Scrapling · MCP · Selenium
• Docker
• CLI
• Platform support
• Plans and concurrency
• FAQ
• License

Why antibrow

The spoofing is in the engine, not in a script. Most stealth tooling patches JavaScript from the outside: override a getter, shim navigator, monkey-patch toString. Anti-bot vendors have been fingerprinting those patches for years. antibrow ships a modified Chromium — Canvas, WebGL, WebGPU, audio, fonts, navigator, screen, DOMRect and timezone are answered inside C++/Blink, so there is no injected script to find, no property descriptor out of place, and worker contexts return exactly what the main thread does.

One coherent identity, frozen per profile. Randomising every value independently is itself a signal: real devices do not pair an AMD renderer with an Intel vendor string, or a 1.0 device pixel ratio with a 1536×864 screen. Each profile gets one self-consistent persona at creation time and keeps it forever — same UA, same GPU, same seeds, same fonts, launch after launch.

Timezone follows the proxy. Pass a proxy and the exit IP is resolved through that same proxy; the browser's timezone and WebRTC identity are set from it before the first byte of the first page.

Unlimited local profiles, free. A profile is a directory. Name one and it exists. There is no per-profile fee and nothing to provision. Your plan sets how many browsers run at the same time, not how many identities you may own.

Built for agents. launch() hands back a live CDP endpoint plus a Playwright handle, so browser-use, crawl4ai, Scrapling, your own MCP server, or plain Playwright all attach without glue code.

Install

pip install antibrow

Then fetch the browser kernel and store your API key (both are one-time):

python -m antibrow install     # downloads + extracts the kernel (~190 MB zip, ~440 MB on disk)
python -m antibrow login       # stores your key in ~/.antibrow/license.key

An API key is required. The kernel verifies a short-lived, server-signed license token on startup and refuses to run without one — that check is compiled into the binary, so there is no offline mode. A free key (1 concurrent browser, unlimited local profiles) is at antibrow.com. This package never signs tokens itself: it exchanges your key for a token over HTTPS and caches it, so a tight relaunch loop hits the network roughly once a day.

You do not need playwright install — antibrow drives its own kernel, not Playwright's bundled browsers. The playwright package is still required (for its client library).

Skipping install is fine: the first launch() downloads the kernel it needs.

Quick start

from antibrow import launch

# A named profile: same fingerprint, cookies and storage every time.
browser = launch(profile="shopper-01")

page = browser.new_page()
page.goto("https://whoer.net")
print(page.title())

browser.close()

Context manager, headless, proxy, geo-matched timezone:

from antibrow import launch

with launch(
    profile="scraper-eu",
    headless=True,
    proxy="http://user:[e-mail]:8080",
    geoip=True,                       # timezone + WebRTC follow the proxy exit
) as browser:
    page = browser.new_page()
    page.goto("https://example.com")
    print(browser.timezone, browser.public_ip)

Async, for agents and concurrent crawls:

import asyncio
from antibrow import launch_async

async def main():
    browser = await launch_async(profile="agent-01")
    page = await browser.new_page()
    await page.goto("https://example.com")
    print(await page.title())
    await browser.close()

asyncio.run(main())

More in examples/.

API reference

launch(profile="default", **options) -> Antibrow

Starts the kernel and returns a handle that is ready to drive. Blocking (sync) API.

Option	Type	Default	What it does
profile	str	"default"	Profile name. Same name → same identity, cookies, storage. Unlimited, free, local.
headless	bool	False	Hide the window. On Windows the window is moved off-screen instead of --headless=new, because headless Chromium has its own detectable fingerprint. On Linux use Xvfb (see Docker).
proxy	str | dict	None	"http://user:pass@host:port", "socks5://…", "https://…", "relay://…", or Playwright's {"server": …, "username": …, "password": …}.
geoip	bool	True	Resolve the proxy's exit IP through the proxy and make timezone + WebRTC match it. No-op without a proxy.
timezone	str	None	Force an IANA timezone ("Europe/Berlin"), overriding the geo lookup.
api_key	str	env / key file	AntiBrow API key.
server	str	https://antibrow.com	License server base URL.
cache_dir	path	~/.anti-detect-browser	Where kernels and profiles live.
profile_dir	path	None	Exact profile directory, bypassing cache_dir/profile.
kernel_version	str	newest	Kernel for a new profile. Existing profiles keep the version frozen in their persona.
label	str	profile name	Text shown in the kernel's address-bar tag — tells windows apart at a glance.
args	list[str]	None	Extra Chromium switches.
proxy_auth	"native" | "extension"	"native"	How proxy credentials are answered. Native = inside the network stack, no extension.
license_token	str	None	Use a pre-minted token instead of calling the server.
license_provider	callable	None	Return a token from your own issuer (self-hosted, vault, CI).
update_kernel	bool	False	Check for a newer build of this profile's kernel and install it before launching.
reuse_initial_page	bool	True	Let the first new_page() return Chromium's initial blank tab instead of opening a second one.
timeout	float	120.0	Seconds to wait for the browser to come up.
on_progress	callable	None	Receives progress lines ("Downloading 42%", "CDP endpoint ready …").

The Antibrow handle

Attribute lookups fall through to the Playwright BrowserContext, so anything you would call on a context works directly on the handle.

browser = launch(profile="p1")

page  = browser.new_page()          # -> Playwright Page
pages = browser.pages               # -> delegated to the context
browser.add_init_script("…")        # -> delegated
browser.add_cookies([...])          # -> delegated

browser.context                     # the raw Playwright BrowserContext
browser.browser                     # the raw Playwright Browser (CDP connection)
browser.page                        # first page, created on demand

browser.cdp_endpoint                # 'ws://127.0.0.1:54321/devtools/browser/…'
browser.cdp_url                     # 'http://127.0.0.1:54321'  (what crawl4ai wants)
browser.profile_dir                 # Path to this profile on disk
browser.persona                     # the frozen identity (UA, GPU, screen, seeds…)
browser.timezone, browser.public_ip # resolved from the proxy when geoip=True
browser.kernel_version, browser.pid
browser.plan                        # everything resolved for this launch
browser.plan.redacted_args()        # the command line, secrets masked - paste into bug reports

browser.close()                     # closes the browser and reaps the process tree

new_page() hands back Chromium's initial blank tab the first time it is called (Chromium always opens with one), then opens real new tabs. Use browser.context.new_page() if you always want a fresh one.

Other entry points

from antibrow import launch_async, launch_persistent_context, prepare_launch

browser = await launch_async(profile="p1")          # asyncio twin of launch()
context = launch_persistent_context(profile="p1")   # raw Playwright BrowserContext
plan    = prepare_launch(profile="p1")              # resolve everything, start nothing

prepare_launch() returns the exact executable, arguments, persona and timezone that a launch would use, without starting a process — useful for tests, dry runs and bug reports.

Errors

Every intentional failure derives from AntibrowError:

from antibrow import AntibrowError, ConcurrencyLimitError, LicenseError

try:
    browser = launch()
except ConcurrencyLimitError:
    ...   # the plan's simultaneous-browser cap is in use (enforced by the kernel)
except LicenseError:
    ...   # no API key, or the server rejected it
except AntibrowError:
    ...   # kernel download, unsupported platform, proxy, launch failure

Profiles and fingerprints

A profile is a directory under ~/.anti-detect-browser/profiles/<name>/:

persona.json     the frozen identity - written once, never regenerated
fp-config.json   the persona serialized for the kernel, rewritten each launch
user-data/       Chromium's profile: cookies, storage, history, extensions

The cache directory is shared with the Node SDK and the AntiBrow desktop app, so a profile created from Python shows up in the desktop app's list and vice versa. Override it with ANTIBROW_CACHE_DIR or cache_dir=.

What a persona pins:

Surface	Example
User agent + navigator	Windows 11 / Chrome 150, platform, vendor, hardwareConcurrency, deviceMemory, maxTouchPoints, UA-CH platformVersion
Screen	CSS size, availWidth/Height minus the taskbar, colorDepth, devicePixelRatio (never 1.0)
GPU	matching WebGL unmasked vendor + renderer (Intel / NVIDIA / AMD)
Canvas, audio, DOMRect	per-profile seeds → deterministic noise, identical on every visit
Fonts	Windows font set, Segoe UI, no CJK leakage
Locale	languages, Accept-Language, and the timezone (from the proxy when geoip=True)
WebRTC	passthrough with the proxy's public IP, or disabled when there is no proxy

Determinism matters as much as the values themselves: a browser that returns a new canvas hash on every call is trivially flagged. Seeds are stable per profile, so repeat visits agree with each other.

Inspect a live identity:

browser = launch(profile="p1")
print(browser.persona.ua, browser.persona.gpu_renderer, browser.persona.screen_w)

Sanity checks worth running once: creepjs, whoer.net, browserleaks.com/canvas, pixelscan.net.

Proxies

launch(proxy="http://user:[e-mail]:8080")
launch(proxy="https://user:[e-mail]:443")
launch(proxy="socks5://user:pass@127.0.0.1:1080")
launch(proxy={"server": "http://gate.example.com:8080", "username": "u", "password": "p"})

Credentials are handled inside the kernel — HTTP/HTTPS 407 challenges are answered in the network stack, SOCKS5 uses RFC 1929 user/password negotiation. Nothing is loaded into chrome://extensions, which is exactly the kind of tell an antidetect browser must not have. (proxy_auth="extension" reproduces the older MV3 approach if you ever need it for an HTTP proxy; it cannot do SOCKS5.)

Passwords containing @, : or / are fine — percent-encode them in the URL, or use the dict form.

With geoip=True (the default), the exit IP is looked up through the proxy before launch, and its timezone is written into the fingerprint:

browser = launch(profile="p1", proxy="socks5://user:pass@127.0.0.1:1080")
print(browser.public_ip, browser.timezone)   # 203.0.113.7 America/Los_Angeles

AntiBrow's own managed residential proxies use the relay:// scheme, which the kernel speaks natively:

launch(proxy="relay://<api-key>:<proxy-id>@proxy.antibrow.com")

Framework integrations

Every integration works the same way: antibrow starts the browser, and you hand its CDP endpoint to whatever wants to drive it.

browser = launch(profile="p1")
browser.cdp_url        # http://127.0.0.1:54321
browser.cdp_endpoint   # ws://127.0.0.1:54321/devtools/browser/…

Playwright

The handle is Playwright. Existing scripts change only their launch line:

# before
# from playwright.sync_api import sync_playwright
# pw = sync_playwright().start()
# browser = pw.chromium.launch()
# context = browser.new_context()

from antibrow import launch
context = launch(profile="p1")        # a BrowserContext in all but name

page = context.new_page()
page.goto("https://example.com")
page.get_by_role("button", name="Sign in").click()

Need the literal object for an API that type-checks it:

from antibrow import launch_persistent_context
context = launch_persistent_context(profile="p1")   # playwright BrowserContext
context.close()                                     # also stops the kernel

Full example: examples/04_playwright.py.

Coming from Puppeteer or the Node SDK

Same product, two runtimes — anti-detect-browser on npm, antibrow on PyPI, sharing one cache directory, one profile format and one account.

// Node
const ab = new AntiDetectBrowser({ key: process.env.ANTI_DETECT_BROWSER_KEY })
const { page, browser } = await ab.launch({ profile: 'shopper-01' })
await page.goto('https://example.com')
await browser.close()

# Python
browser = launch(profile="shopper-01")
page = browser.new_page()
page.goto("https://example.com")
browser.close()

Puppeteer users: the endpoint is plain CDP, so puppeteer.connect({ browserURL: browser.cdp_url }) works from any language. Side-by-side mapping in examples/05_puppeteer_style.py.

browser-use

from antibrow import launch_async
from browser_use import Agent, Browser, ChatOpenAI

session = await launch_async(profile="agent-01", proxy="http://user:pass@gate:8080")
agent = Agent(
    task="Find the cheapest flight from Berlin to Lisbon next month",
    llm=ChatOpenAI(model="gpt-4.1-mini"),
    browser=Browser(cdp_url=session.cdp_url),
)
await agent.run()

examples/06_browser_use.py — includes the fallback spelling for older browser-use releases (BrowserSession(cdp_url=…)).

crawl4ai

from antibrow import launch_async
from crawl4ai import AsyncWebCrawler, BrowserConfig

session = await launch_async(profile="crawler-01")
config = BrowserConfig(cdp_url=session.cdp_url, headless=False)

async with AsyncWebCrawler(config=config) as crawler:
    result = await crawler.arun(url="https://example.com")
    print(result.markdown)

examples/07_crawl4ai.py

Scrapling

from antibrow import launch
from scrapling.fetchers import DynamicFetcher

browser = launch(profile="scrapling-01")
page = DynamicFetcher.fetch("https://example.com", cdp_url=browser.cdp_endpoint)
print(page.css_first("h1::text"))

examples/08_scrapling.py

AI agents and MCP

Any MCP client can drive a fingerprinted browser. examples/09_mcp_server.py is a complete stdio MCP server (pip install "antibrow[mcp]") exposing launch_browser, navigate, click, fill, get_content, screenshot, evaluate and close_browser:

{
  "mcpServers": {
    "antibrow": {
      "command": "python",
      "args": ["/abs/path/to/examples/09_mcp_server.py"],
      "env": { "ANTIBROW_API_KEY": "your-api-key" }
    }
  }
}

The Node package ships an MCP server out of the box (npx anti-detect-browser --mcp) if you would rather not run the example.

Selenium

Selenium cannot attach to a CDP-only endpoint without a matching chromedriver, so there is no supported Selenium binding today. If you are migrating, the Playwright section above is the shortest path; open an issue if chromedriver support matters to you.

Docker

The Linux kernel runs headful under Xvfb — real headless Chromium has its own fingerprint, so the image renders to a virtual display instead.

FROM python:3.12-slim
RUN apt-get update && apt-get install -y --no-install-recommends \
      xvfb libnss3 libatk1.0-0 libatk-bridge2.0-0 libcups2 libdrm2 \
      libxkbcommon0 libxcomposite1 libxdamage1 libxfixes3 libxrandr2 \
      libgbm1 libasound2 libpango-1.0-0 libcairo2 fonts-liberation ca-certificates \
    && rm -rf /var/lib/apt/lists/*
RUN pip install --no-cache-dir antibrow
COPY script.py .
CMD ["xvfb-run", "-a", "python", "script.py"]

docker build -t my-scraper .
docker run --rm -e ANTIBROW_API_KEY=$ANTIBROW_API_KEY \
  -v antibrow-cache:/root/.anti-detect-browser my-scraper

Mounting the cache volume keeps the kernel (and your profiles) between runs. The full image is in this repo's Dockerfile; see examples/10_docker/.

CLI

python -m antibrow install [--version 150.0.7871.182] [--force]   # get the kernel
python -m antibrow info                                           # kernels, profiles, license
python -m antibrow login [--key ab_live_…]                        # store an API key
python -m antibrow version                                        # SDK + default kernel

antibrow … works too (console script). info is the first thing to run when something is wrong: it prints the cache directory, every kernel version with install/update status, all profiles with their pinned kernel, and where your API key was found.

Environment variables

Variable	Purpose
ANTIBROW_API_KEY	API key (also accepts the Node SDK's ANTI_DETECT_BROWSER_KEY)
ANTIBROW_LICENSE_TOKEN	Pre-minted license token; skips the server call entirely
ANTIBROW_CACHE_DIR	Kernel + profile root (default ~/.anti-detect-browser)
ANTIBROW_SERVER	License server base URL

Platform support

Platform	Status	Notes
Windows 10/11 x64	Supported	Headful, or headless via off-screen window
Linux x64 (glibc)	Supported	Needs Xvfb for headless; container flags applied automatically
Docker (linux/amd64)	Supported	See Docker
macOS	Not yet	No kernel build; launch() raises UnsupportedPlatformError
Linux arm64 / musl	Not yet	No kernel build

Python 3.9 – 3.13. The kernel is a ~190 MB download (~440 MB extracted), cached once per version — python -m antibrow info shows what is installed and where.

Plans and concurrency

Local profiles are unlimited on every plan, including free. What scales with the plan is how many browsers run at the same time — enforced by the kernel itself (cross-process file locks), not by this SDK, so it cannot be worked around by spawning more Python processes.

Plan	Local profiles	Concurrent browsers	Cloud sync	Managed proxies
Free	unlimited	1	–	–
Basic	unlimited	5	yes	yes
Pro	unlimited	20	yes	yes
Team	unlimited	100	yes	yes

Details at antibrow.com/pricing. Exceeding the cap raises ConcurrencyLimitError instead of hanging.

Cloud profile sync and Live View are implemented in the Node SDK and the desktop app; this package is local-only for now.

FAQ

Do I need playwright install? No. antibrow downloads and drives its own kernel. The playwright pip package is required for its client library, but its bundled browsers are never used.

Does it work without an API key? No. The license check is compiled into the kernel binary. A free key is at antibrow.com, and one token covers a whole day of relaunches.

Where does my data go? Profiles never leave your machine in this package. The only outbound calls are: the kernel download (download.antibrow.com), the token exchange (antibrow.com), and — only when geoip=True and a proxy is set — one request to ip-api.com through your proxy to read its exit timezone.

Can I use my own profile directory / mount it into CI? Yes: launch(profile_dir="/data/profiles/acct-17"), or set ANTIBROW_CACHE_DIR. Copy the directory to move an identity between machines.

Is headless detectable? Real headless Chromium is, which is why headless=True on Windows moves the window off-screen instead. On Linux, run headful under Xvfb (as the Dockerfile does).

How do I keep the kernel up to date? python -m antibrow install --force, or launch(update_kernel=True). Installed kernels are never swapped under you.

Something fails on launch — what do I send? python -m antibrow info and prepare_launch(...).redacted_args(). Both are safe to paste; the license token and proxy password are masked.

Is this legal to use? It is a browser. Scraping public data, testing your own anti-fraud stack and managing your own accounts are ordinary uses. Fraud, credential stuffing and violating a site's terms are not — and are not supported here.

License

Two licenses, and the boundary matters:

• This repository — the Python wrapper, CLI, examples and docs — is MIT. Fork it, vendor it, ship it.
• The browser kernel binary is closed source and separately licensed. It is not in this repository and not in the PyPI package; it is downloaded from AntiBrow's own CDN at runtime, by the end user, onto the end user's machine. Redistributing, reselling or repackaging that binary is not permitted. Full terms and the OEM/SaaS boundary: BINARY-LICENSE.md.

Depending on this package does not make you a redistributor of the kernel.

Links

• Website and API keys — antibrow.com
• Docs — antibrow.com/docs/sdk
• Node/TypeScript SDK — anti-detect-browser
• Desktop app — antibrow.com/download
• Issues — GitHub