schauto

Stealth Chromium automation framework with kernel-level fingerprint randomization. Built by patching Chromium source directly — no JS-level overrides that can be detected by Object.getOwnPropertyDescriptor.

Features

• Fingerprint randomization at C++ level via FingerprintToolkit (per-context xorshift64 seed)
  • Canvas / WebGL image data noise
  • AudioContext buffer perturbation
  • ClientRects coordinate jitter
  • WebGL vendor / renderer spoofing
  • Navigator hardwareConcurrency, platform
  • SpeechSynthesis voices filtering
  • Timezone spoofing
  • Font allowlist filtering
  • WebRTC IP leak prevention (proxy-bound)
• No-window mode (kernel-level, not just --headless flag)
• Auto geo-match from proxy IP (timezone, language, platform via ip-api.com → ipwho.is → ipapi.co fallback chain)
• Built-in AdBlock Plus MV3 extension (auto-loaded)
• SOCKS5 / HTTP proxy support
• Bezier mouse curves with jitter for captcha solving
• Image API: full-page screenshot, region screenshot, element screenshot, boundingBox

Installation

npm install schauto

The package ships with the patched Chromium binary (~400MB). First install will take a while.

Usage

import { Browser } from 'schauto'

const browser = await Browser.launch({
  proxy: 'socks5://user:pass@proxy.example.com:1080',
  // or 'http://127.0.0.1:7890'
})

const ctx = await browser.newContext()
const page = await ctx.newPage()

await page.goto('https://bot.sannysoft.com/')

const png = await page.image.screenshot()
require('fs').writeFileSync('result.png', png)

// Mouse with Bezier curve + jitter (anti-captcha)
await page.mouse.move(100, 100, { x: 500, y: 400 })
await page.mouse.click(500, 400)

// Element screenshot
const cropped = await page.image.screenshotElement('#captcha-image')

await browser.close()

Architecture

schauto/
├── src/                    TS source (Browser, BrowserContext, Page, Mouse, ImageHelper, CDP client)
├── chromium-patches/       FingerprintToolkit C++ source (copied into Chromium tree at build time)
├── patches/                Unified diffs applied to Chromium source
│   ├── 00-local-frame-toolkit.patch    LocalFrame member injection
│   ├── 00b-frame-init-toolkit.patch    LocalFrame::Init() reads CLI flags
│   ├── 01-no-window.patch              No-window mode
│   ├── 02-canvas-noise.patch           Canvas data noise
│   ├── 03-webgl-spoof.patch            WebGL VENDOR/RENDERER override
│   ├── 04-audio-noise.patch            AudioContext buffer noise
│   ├── 05-client-rects.patch           ClientRects coordinate jitter
│   ├── 06-navigator.patch              Navigator API spoofing
│   ├── 07-webrtc-ip.patch              WebRTC IP leak prevention
│   ├── 08-timezone.patch               Timezone spoofing
│   ├── 09-speech-voices.patch          SpeechSynthesis voices
│   └── 10-font-filter.patch            Font allowlist
├── build/                  Build scripts
│   ├── fetch-chromium.sh   Fetches Chromium source via gclient
│   ├── apply-patches.py    Applies patches programmatically
│   ├── apply-patches.sh    Wrapper
│   └── args.gn             Chromium build args (Release, no-symbols, no-component, etc.)
├── extensions/
│   └── adblock-plus/       MV3 extension auto-loaded into every context
└── dist/                   Compiled JS (output of `npm run build`)

Building from source

Requires:

• Windows 10/11 with Visual Studio 2019+ Build Tools
• depot_tools from Chromium project
• ~150GB disk space, 16GB+ RAM, several hours

# 1. Fetch Chromium (slow, ~80GB)
./build/fetch-chromium.sh

# 2. Apply patches
python build/apply-patches.py

# 3. Configure GN
cd chromium-src/src
cp ../../build/args.gn out/Release/args.gn
gn gen out/Release

# 4. Compile (4-8 hours)
ninja -C out/Release chrome

# 5. Build npm package
cd ../..
npm run build

API

Browser.launch(opts: LaunchOptions): Promise<Browser>

interface LaunchOptions {
  proxy?: string              // 'socks5://...' or 'http://...'
  executablePath?: string     // Override chrome.exe path
}

BrowserContext

• newPage(): Promise<Page>
• close(): Promise<void>

Page

• goto(url: string): Promise<void>
• evaluate<T>(fn: string | Function, ...args): Promise<T>
• $(selector): Promise<string | null>
• screenshot(): Promise<Buffer>
• close(): Promise<void>
• mouse: Mouse — Bezier-curve mouse helper
• image: ImageHelper — screenshot helpers

ImageHelper

• screenshot(): Promise<Buffer>
• screenshotRegion(x, y, w, h): Promise<Buffer>
• screenshotElement(selector): Promise<Buffer | null>
• boundingBox(selector): Promise<{x, y, width, height} | null>

Mouse

• move(from: {x,y}, to: {x,y}): Promise<void> — Bezier curve with jitter
• click(x, y): Promise<void>

License

MIT