Introducing Precursor: detecting agentic behavior with continuous client-side signals

Summary

Cloudflare has introduced Precursor, a client-side verification system that extends bot detection across an entire user session rather than relying on isolated checkpoints such as login or checkout pages. The service injects a compact, dynamically assembled JavaScript bundle into HTML responses to gather lightweight behavioral signals—pointer trajectories, keyboard timing and rhythm, focus shifts, and page visibility—without capturing raw keystrokes or linking data to user identities. These signals are buffered locally and forwarded at intervals to Cloudflare’s edge for real-time evaluation.

Precursor addresses the limitation of earlier tools like Turnstile, which already handles billions of managed challenges daily but leaves most of an application’s interaction surface unmonitored. Modern automation can execute JavaScript in real browsers and pass single challenges convincingly; consistent human-like behavior sustained over minutes proves far harder to replicate. Evaluators examine correlations such as pointer activity matching visible page time or keyboard events occurring only while a field is focused, then feed consolidated signals into the existing bot-scoring pipeline.

Because Precursor maintains a session-scoped view, a bot cannot reset its behavioral profile by reloading a page. The system also surfaces session-level analytics that let operators inspect typical journeys, identify divergence from expected patterns, and observe automation indicators that appear only across multiple requests. It is offered as an optional Enterprise Bot Management feature and runs initially without additional friction while the operator tunes enforcement thresholds.

Why it matters

This article is highly relevant for security and privacy professionals in the Netherlands as it addresses the growing threat of AI agents bypassing traditional security measures. Its 'privacy by design' approach aligns well with strict EU/GDPR requirements, offering a compliant way to secure applications against advanced automation.