jo-inc

camofox-browser

AI#AI Agents#Browser Automation#Anti-detection#Playwright#Web Scraping

// summary

Camofox-browser is a specialized server designed to provide AI agents with reliable web browsing capabilities by leveraging the Camoufox engine for C++ level fingerprint spoofing. It offers a REST API that simplifies interactions through accessibility snapshots, stable element references, and built-in search macros. The system is optimized for efficiency and deployment, featuring automatic idle shutdown and session isolation to support scalable agent operations.

// technical analysis

Camofox-browser is an anti-detection browser server designed specifically for AI agents, leveraging a custom Firefox fork that implements fingerprint spoofing at the C++ level. By addressing the limitations of standard automation tools like Playwright and headless Chrome, it provides a robust, stealthy environment that bypasses modern bot detection mechanisms. The project prioritizes efficiency and reliability for autonomous agents through features like accessibility snapshots, stable element references, and intelligent session management, making it suitable for resource-constrained environments like VPS or shared infrastructure.

// key highlights

Implements C++ level anti-detection to effectively bypass sophisticated bot protection systems like Cloudflare and Google.

Provides stable element references (e1, e2, etc.) to ensure reliable interaction for AI agents regardless of DOM changes.

Generates token-efficient accessibility snapshots that are approximately 90% smaller than raw HTML, optimizing LLM context usage.

Features a lightweight architecture that maintains a ~40MB memory footprint when idle, allowing for deployment on low-cost hardware.

Supports session isolation and cookie injection, enabling authenticated browsing sessions without requiring interactive logins.

Includes built-in search macros and automated GeoIP/proxy configuration to simplify complex web navigation and data retrieval tasks.

// use cases

Bypassing advanced bot detection on sites like Google and Cloudflare using C++ level fingerprint spoofing

Enabling AI agents to interact with the web via token-efficient accessibility snapshots and stable element refs

Managing authenticated browsing sessions through automated cookie injection and proxy-based GeoIP routing

// getting started

To begin, clone the repository and run 'npm install' followed by 'npm start' to launch the server on port 9377. For containerized environments, use the provided Makefile to build and deploy via Docker, ensuring the necessary browser binaries are fetched first. Developers can then interact with the browser via the REST API or by installing the OpenClaw plugin.