---
title: "Scalebrowser Blog, page 6"
description: "Guides and measurements for teams whose AI agents drive a real browser: detection, captchas, sign-ins and infrastructure."
canonical: "https://scalebrowser.net/blog/page/6"
---

> ## Documentation Index
> Fetch the complete documentation index at: https://scalebrowser.net/llms.txt
> Use this file to discover all available pages before exploring further.

# Scalebrowser Blog, page 6

> Guides and measurements for teams whose AI agents drive a real browser: detection, captchas, sign-ins and infrastructure.

- [Canvas fingerprinting: why profiles on one machine match](https://scalebrowser.net/blog/canvas-fingerprinting.md): What canvas fingerprinting reads, which parts of a machine decide the hash, and why browser profiles on one computer share it without looking fake.
- [Captcha solving without a service: 18 types measured](https://scalebrowser.net/blog/captcha-solving.md): 18 challenge types an AI agent solved in its own browser on live widgets with real keys, no solving service, with run counts and the 2 that stay open.
- [Chrome DevTools MCP: 393,149 tokens per Wikipedia page](https://scalebrowser.net/blog/chrome-devtools-mcp.md): Chrome DevTools MCP measured as an agent's browser: tokens per page and per step on 3 live pages, the context window it fills and what it is built for.
- [Is CloakBrowser good for AI agents? Fix the seed and noise](https://scalebrowser.net/blog/cloakbrowser.md): CloakBrowser reviewed: what its 87 Chromium patches cover, its 26 of 31 benchmark result, the free and Pro builds, and which 2 defaults to change on day one.
- [Cloudflare Turnstile: can an AI agent pass it?](https://scalebrowser.net/blog/cloudflare-turnstile.md): Cloudflare Turnstile measured with an AI agent on 3 live pages with production keys: tokens in 1.7 and 2.9 seconds, one variant that failed twice, and why.
- [hCaptcha and AI agents: 4 of 4 solved, 0 on its own demo](https://scalebrowser.net/blog/hcaptcha.md): How an AI agent solved hCaptcha's growing-shape task by measuring frames, 4 of 4 on one site, and why hCaptcha's own demo refused the same correct answers.
- [JA4 fingerprint: 1 TLS extension tells Chrome from a fork](https://scalebrowser.net/blog/ja4-tls-fingerprint.md): What the JA4 TLS fingerprint reads before any script runs, why JA3 cannot follow Chrome, and the one extension that split our Chromium build from real Chrome.
- [nodriver vs undetected-chromedriver: which one to use](https://scalebrowser.net/blog/nodriver.md): nodriver against undetected-chromedriver: releases, licences, a 31 target benchmark, and the 3 things nodriver leaves to your code: profile, input, licence.
- [Patchright: Playwright minus its CDP leaks, and what remains](https://scalebrowser.net/blog/patchright.md): Patchright swaps one import and hides Playwright's Runtime.enable and Console.enable leaks. What a detector still sees, and its score in a 31-site benchmark.
- [Playwright MCP token usage: 604,785 tokens to see 1 page](https://scalebrowser.net/blog/playwright-mcp-tokens.md): Playwright MCP measured in tokens per page and per step against 5 other browser tools, with the cost curve of a session and what cuts it.
- [Playwright stealth for Python and Node: what it patches](https://scalebrowser.net/blog/playwright-stealth.md): What playwright-stealth (Python) and playwright-extra (Node) patch, what a detector still sees, and what Playwright's own typing looks like to a page.
- [Is puppeteer-extra-plugin-stealth still maintained?](https://scalebrowser.net/blog/puppeteer-stealth.md): The Puppeteer stealth plugin read on 22 September 2026: maintenance, what its 17 evasions change, where they stop, and which maintained tools replace it.

Page 6 of 7. Previous: [page 5](https://scalebrowser.net/blog/page/5.md). Next: [page 7](https://scalebrowser.net/blog/page/7.md).