Camofox: Stealth Browser for AI Agents

Why Camofox now

AI agents that actually crawl the web need browser automation, but existing tools keep getting blocked. Cloudflare blocks, headless browser fingerprints, anti-bot plugins that keep changing — Camofox steps in at the browser engine level to solve these.

Changed core issues

Camofox is built on two layers: the browser engine and the control API. The main engine, Camoufox, is based on Firefox, but with C++-level performance improvements. Before randomizing, core signals like navigator.hardwareConcurrency, WebGL renderer, AudioContext, screen resolution, and WebRTC values are already modified. The noise resistance is higher than simple shims because it operates at the engine level. On top of that, a Node.js REST API lets agents control the browser programmatically. Stable element references (e1, e2, e3) and an accessibility-based traversal layer are key. The traversal layer creates about 90% fewer tokens than the original HTML, saving tokens, and click/input are handled stably by reference ID.

Key features and operational characteristics

As of August 2026, v1.14.0 has been released and features have stabilized. It includes over 10 built-in search macros like @google_search, @youtube_search, @amazon_search, @reddit_subreddit, so sites often work without separate crawling. Session persistence, cookie import, file upload, proxy/GeoIP routing, YouTube caption extraction, stealth screenshot, large-page open pagination, download detection, DOM image extraction, and other automation essentials are doubled. Request IDs can be dropped from JSON structured logs for operational stability, and Docker/Fly.io/Railway deployment settings are also supported. It can also be deployed as an OpenClaw plugin; in OpenClaw config, use openclaw plugins install @askjo/camofox-browser, and MCP server mode is also available for MCP-compatible clients.

How far can we use it

The biggest advantage over competing tools is the detection evasion rate. The stealth success rate reflects the user's proxy, session quality, and usage patterns, and there is no absolute guarantee. Also, since Camoufox is Firefox-based, compatibility with Chrome-first web standards and some extensions may have issues. Open source telemetry is basically off by default. Individual builds use HMAC, and route/token/IP are excluded from reports, but full opt-in is possible by setting CAMOFOX_CRASH_REPORT_ENABLED=false or CAMOFOX_CRASH_REPORT_URL to a custom endpoint.

Limitations and considerations

The compatibility gap with Chrome-based environments can be a barrier for some projects. If your automation depends on Chrome DevTools Protocol or extension APIs, verify browser support first.

Conclusion

Camofox is a practical choice for AI agent browser automation. With C++-level engine speed improvements, high stealth success rates, and Docker/OpenClaw/MCP connectivity, it's worth considering. For teams that want session stability, proxy, logging, and deployment convenience, Camofox deserves attention.

Reference links

• Camofox GitHub
• Camoufox Official Site
• OpenClaw