Security & Guardrails.
How Fognitix keeps browser automation safe. Our multi-layered security controls prevent prompt-injection and hold critical actions for review.
01. Confirmation Gates
Fognitix is designed to execute tasks autonomously up to 26 actions, but it operates with built-in boundaries. Any irreversible transaction—such as clicking checkout buttons, submitting payments, publishing posts, sending outbound communications, or deleting data files—is paused and held. Fognitix waits for your explicit click to approve before executing.
02. Immediate Emergency Stop
You maintain real-time monitoring of the agent. An omni-present "Stop Agent" override button is accessible in the main toolbar. Clicking this instantly halts the browser executor engine, cuts connection loops, and transfers manual mouse and keyboard controls back to you.
03. Untrusted Page Isolation
Web page content is treated as untrusted and potentially hostile to resist prompt-injection attacks. Accessibility trees and page coordinate outputs are structured in sandboxed namespaces. This prevents raw script contents on web pages from hijacking the agent model's instructions.
04. Proxied API Connections
All browser automation requests and AI model queries are proxied. This ensures that direct authorization keys, host configurations, and critical cookies never interface with the outer web layer, protecting your workspace integrity.