🤖 Anti-bot tells
WebDriver / CDP / Puppeteer / Playwright / Selenium / headless artefacts.
…
trust
Running checks…
Spawning ~40 detection probes — Canvas, WebGL, Audio, AT-SPI markers, TLS, IP rep…
🕵️ Anti-anti-detect patches
Tells left by Camoufox / undetected-chromedriver / stealth plugins — toString() tampering, missing-API stubs, juggler stack frames, Firefox-on-Windows mismatches.
🎯 Fingerprint consistency
UA vs userAgentData, languages vs Accept-Language, timezone vs IP geo, screen vs window vs DPR, hardware claims.
🖥️ Remote / virtual display
VNC / AnyDesk / RustDesk / Xvfb / headless Chrome remote-debugging tells — software GPU rasterizer (llvmpipe), forced colour-depth, integer-only DPR, missing touch / pressure, quantized timer, virtual audio devices.
🌐 Network / IP
ASN, datacenter / proxy / VPN / Tor classification, WebRTC IP leak, DNS leak indicators.
🖼️ Fingerprint hashes
Stable signatures the anti-bot grader compares against known-bad sets. Should change with every fresh profile but stay stable across page reloads of the same session.
📋 Raw environment
UA, platform, plugins, languages, screen — the values themselves, for the operator's eye.
How to read this: trust = 100 − Σ severity of failed probes. Anti-bot graders weight catastrophic tells (e.g. navigator.webdriver = true) heavily and consistency tells (e.g. UA says Windows but platform says Linux) almost as heavily. A real human residential profile scores 90+; a stock Selenium scores 0–20; a well-stealthed bot scores 70–95 depending on which patches it ships.
Reload to re-run. Open with the operator's regular Mac browser as a reference, then with one of our agent's Camoufox to compare side-by-side.