Enable AI agents to browse the web without triggering bot detection
This tool stops AI agents from getting stuck on 'Are you a robot?' prompts. It spoofs browser fingerprint information at the deepest level, allowing agents to navigate the web as naturally as a human user. You can connect your existing automation code without complex configuration.
How it changes fingerprints inside the browser
Existing stealth plugins often try to change values at the JavaScript level, which can actually provide clues for bot detection. This project is based on Camoufox, a Firefox fork, and spoofs hardware information, WebGL renderer, audio context, screen size, and WebRTC data at the C++ implementation level. All fingerprint information is manipulated before JavaScript executes, leaving no detectable traces from the outside. This allows it to bypass Google, Cloudflare, and most bot detection systems.
Page structure that is easy for agents to read
It is inefficient for agents to process huge HTML code to understand web pages. This server converts pages into accessibility snapshots. This format is approximately 90% smaller than the original HTML, allowing agents to quickly grasp only the information they need. It also assigns stable reference numbers like e1 and e2 to each element, ensuring that agents can accurately manipulate the desired element even if its position changes. This structure plays a key role in reducing token usage and increasing interaction reliability.
Lightweight execution environment and session management
Keeping a browser running constantly consumes a lot of memory, but this tool automatically terminates the browser when it is idle, keeping memory usage at approximately 40MB. This allows it to run stably alongside other services in restricted environments such as Raspberry Pi or low-cost VPS. It isolates sessions by separating cookies and storage per user, and you can import Netscape-format cookie files to visit sites in an authenticated state. Additionally, when routing traffic through a proxy, it automatically sets the locale and timezone to match the IP address, preventing blocks caused by geographic inconsistencies.
Things to check before adoption
This tool is designed to solve the problem of AI agents being blocked from web access by bot detection. It is particularly useful when encountering strong defense walls like Cloudflare in large-scale web scraping or automation tasks. Before adopting it, you should note that downloading the Camoufox binary (approximately 300MB) is required, and that yt-dlp must be installed additionally to use the YouTube subtitle extraction feature. Telemetry is enabled by default, but you can disable it by setting the CAMOFOX_CRASH_REPORT_ENABLED=false environment variable, so you must check this setting in environments where privacy policy is important.
