AIAny
AI Infra2026
Icon for item

CloakBrowser

A Chromium binary patched at the C++ level to evade bot-detection and serve as a drop-in Playwright/Puppeteer replacement. Notable features: source-level fingerprint patches, human-like input emulation, auto-updating binaries, and integrations for Python/Node.js and Docker — useful for scraping, agent-driven browsing, and stealth automation.

Introduction

Most browser-stealth projects tweak flags or inject JS, which detection systems can fingerprint back. CloakBrowser takes a different approach: it modifies Chromium at the C++ source level so the resulting binary reports coherent, real-looking fingerprints to servers. That makes many surface-signature checks (navigator fields, canvas/WebGL hashes, CDP timing) behave like a normal Chrome install without fragile runtime injections.

What Sets It Apart
  • Source-level patches, not JS/flag hacks — the project applies dozens of C++ patches that alter canvas, WebGL, audio, fonts, WebRTC, GPU reporting, CDP input behavior and other signals. This reduces the class of detection signals that arise from runtime patching and flag fiddling. (README lists 49–57 patches across versions.)
  • Drop-in Playwright/Puppeteer replacement — the wrapper exposes familiar APIs for Python and Node.js so most Playwright code runs unchanged; it also supports Puppeteer with caveats around CDP noise.
  • Behavioral humanization and profile features — a one-flag humanize mode simulates Bézier mouse curves, per-character typing, realistic scrolls and holds; persistent profile support, fingerprint seeds, and timezone/locale auto-detection from proxies help maintain consistent device identity.
  • Practical verification — tests in the README report passing Cloudflare Turnstile, FingerprintJS, BrowserScan and a server-verified reCAPTCHA v3 score (~0.9) as of April 2026; latest wrapper/binary noted in README (v0.3.26 / Chromium 146).
Who it's for — and tradeoffs

Great fit if you need stable, repeatable browser fingerprints for scraping, QA, or AI-agent browsing flows and want an API-compatible drop-in for Playwright-based code. It is especially useful when config-level stealth repeatedly breaks after Chromium updates or when you need human-like input timing for behavioral checks.

Look elsewhere if you require guaranteed legal clearance for automation (the README warns against unauthorized abuse) or if you must redistribute the compiled binary freely (the wrapper is MIT but the compiled binary has its own license constraints). Also note: source-level stealth raises an arms-race risk — detection techniques can evolve, so ongoing maintenance is required. macOS Gatekeeper handling and certain platform-specific inconsistencies (documented in the project) may add operational friction.

Where it fits

Use CloakBrowser as infrastructure for large-scale scraping, agent-driven browsing (LangChain/agents that render pages), or integration into Dockerized CI/VPS environments where a reproducible, stealthy Chromium is required. Combine it with residential proxies, persistent profiles, and humanize=True for the highest success rate on aggressive protection stacks.

Practical notes
  • The project provides a prebuilt Docker image and CLI helpers for binary management; binaries auto-download on first run (~200MB).
  • It intentionally focuses on preventing captchas from appearing rather than solving them server-side.
  • Maintain caution: automating sites without authorization can be illegal or violate terms of service; the README explicitly disclaims abusive uses.

Information

  • Websitegithub.com
  • AuthorsCloakHQ
  • Published date2026/02/22

Categories

More Items

GitHub
AI Infra2025

Defines a vendor-neutral JSON/YAML semantic model specification and tooling to exchange metrics, dimensions, lineage and other business semantics across analytics, AI and BI platforms; includes a core spec, validators, converters (dbt, GoodData, Salesforce) and example models.

GitHub
AI Train2025

An asynchronous, high-throughput framework for large-scale reinforcement learning and agentic training that scales to 1T+ MoE models and 1000+ GPUs, with native verifiers integration, end-to-end SFT/RL/evals, and Slurm/Kubernetes deployment; requires NVIDIA GPUs.

GitHub

Runs a self-hosted meeting bot and transcription API that joins Google Meet, Teams and Zoom and streams speaker-attributed transcripts in real time. Compiles meetings into a git-backed Markdown workspace and runs sandboxed agents on your infrastructure; Apache-2.0 and air-gap capable.