Building AI agents inevitably runs into one problem: how to let the agent access web pages.
Not every site has an API. Much of the information still lives in HTML — product prices, competitor data, news content, government filings. Agents need browsers to "see" these pages, but most websites have bot detection.
CloakHQ/CloakBrowser attempts to solve exactly this. 3.4K stars, 30/30 bot detection tests passed, a direct Playwright replacement.
What It Does
The core approach is straightforward: patch browser fingerprints at the source level.
When you use Playwright or Puppeteer to open a page, the browser exposes signals telling the website "I am an automated browser":
navigator.webdriveris true- Chrome headless characteristics
- Canvas fingerprint anomalies
- WebGL rendering differences
- Font list mismatches
CloakBrowser modifies these signals at the Chromium source level, making the browser look like a normal Chrome opened by a user. Not by injecting JavaScript to mask (that is easily detected), but by making the browser "not look like a robot" from the inside.
Test Results
The project claims 30/30 pass rate across bot detection tests. This means in 30 different detection schemes, CloakBrowser was not identified as an automation tool.
This includes Cloudflare Turnstile, reCAPTCHA v2/v3, PerimeterX, Datadome, and other mainstream anti-bot services.
For AI agents doing web data scraping, this is genuinely useful. No need to fumble with fingerprint patching, proxy rotation, or behavior simulation — just use CloakBrowser instead of Playwright.
Production-Ready?
Several points to consider:
License. The project does not have clear commercial licensing terms. If your company is using it, check the LICENSE file carefully.
Maintenance. 151 commits, 48 tags, last update 2 days ago. Activity is decent, but the team is small (primarily the Cloak-HQ account). If a bot detection service updates its logic, the fix speed depends on the author's response time.
Legal risks. Bypassing bot detection is a legal gray area. Different countries have different legal frameworks for web scraping, especially involving user data or trade secrets. Before using CloakBrowser, understand whether the data you are scraping is legally protected.
Connection to AI Agents
This project matters because AI agent web interaction needs are exploding.
Browser-use, Playwright MCP, various browser agent frameworks — all doing the same thing: enabling AI to operate browsers like humans. But if the underlying browser is easily detected, agents cannot do anything.
CloakBrowser fills this bottom-layer gap. It is not an agent framework — it is a tool that lets agents browse "invisibly."
If you are building web interaction features for AI agents, this project belongs on your technology evaluation list. But remember — technical capability and legal compliance are two different things.