Human or bot?
Classified in 8ms.
Real-time traffic classification that separates humans, scripted bots, AI agents, and headless browsers — before they reach your checkout, login, or API.
Not just bots.
Every agent type.
Credential stuffers, scraper bots, AI LLM agents, headless Playwright scripts, and click farms all look different. Serixo classifies each one and tells you exactly what you're dealing with.
Behaviour analysis
Mouse trajectory entropy, scroll physics, click precision, keystroke rhythm, and time distribution across page events — human behaviour is irreproducibly noisy.
40+ behaviour signalsHeadless browser detection
WebDriver flags, Puppeteer/Playwright artifacts, missing browser APIs, inconsistent JS execution timing, and absent media codec support expose automation tools instantly.
Puppeteer · Playwright · SeleniumAI agent classification
LLM-driven agents exhibit distinct request patterns — uniform timing, structured navigation paths, no scroll jitter, and anomalous API call sequences. We classify them separately from simple bots.
GPT · Claude · custom agentsRequest fingerprinting
TLS fingerprint (JA3/JA4), HTTP/2 frame order, header casing, accept-language entropy, and timing jitter between requests create a unique signature for every HTTP client.
JA3 · JA4 · HTTP/2 fingerprintSession velocity
Requests-per-second, page depth per session, checkout abandonment rate, and repeat identical paths are statistical outliers no human produces — even an experienced card tester.
Rate · depth · pattern analysisNetwork reputation
ASN risk scoring, datacenter IP ranges, residential proxy detection, Tor exit nodes, and coordinated IP cluster analysis catch distributed bot farms that rotate IPs to avoid rate limits.
ASN · proxy · Tor detectionEight classes.
One decision.
Serixo doesn't just return “bot” or “human”. Every request receives a precise type label, a confidence score, and a recommended action.
Decision in 8ms.
No CAPTCHA needed.
Signal collection
Browser signals, TLS fingerprint, request headers, and behavioural telemetry are collected passively on every request — no friction, no CAPTCHA, no interruption.
ML classification
A gradient-boosted ensemble model scores 60+ features against 180M+ labelled sessions. Confidence score and type label returned in under 8ms at the edge.
Action triggered
Based on type and confidence, Serixo allows, blocks, challenges, or flags the session. Rules are configurable per endpoint — stricter at checkout, looser on browse.
Risk enrichment
Agent type and confidence flow into Risk Scoring, Device Fingerprint, and Smart Cascade as additional signals — every module gets smarter with detection data.
Tested on
180M+ sessions.
Built to run at the edge with sub-10ms latency. Accurate even on privacy-hardened browsers, residential proxies, and humanised bots.
Stop bots.
Keep humans moving.
One SDK call. Classification on every request. No CAPTCHA, no friction, 0.1% false positive rate.