Browser Agent - Autonomous AI Web Copilot

Next-gen autonomous browser agent: goal-driven planning, spatial vision & native CDP control with local or custom LLMs. 100% BYOK.

Browser Agent is a next-generation autonomous AI browser agent living directly inside your Chrome side panel. Far more than a simple scraping script, it is equipped with a dual-process cognitive hybrid architecture, visual ROI box-selection, demonstration learning, and self-correcting validation. 🚀 Agent Superpowers: 1. ⚡ Dual-Process Cognitive Engine (System 1 + System 2): Sub-second reflexive fast-path (Jev-2B / local model) predicts & executes clicks/inputs via CDP in 200ms–800ms without full-screen screenshot uploads (cutting token usage by 80%+ and boosting speed by 5x–10x); smoothly transitions to System 2 for deep reasoning & verification. 2. 🎯 Visual Region-of-Interest (ROI) Selection: Drag-to-select or snap-click elements via Alt+S for 1:1 pixel-perfect crop, cutting tokens by 80% and eliminating ambiguity. 3. 🔴 Demonstration Learning (Teach-Me): Record actions in the browser; credentials automatically redacted (******); LLM generalizes steps into standard reusable SKILLs. 4. ⏰ Scheduled Skills (Cron): Execute saved workflows periodically (15m–24h) or daily at fixed times in the background with system desktop notifications. 5. 🛡️ Human-in-the-Loop Intervention: Automatically senses sliders/captchas, suspends execution safely, and resumes seamlessly after user solves. 6. ☁️ Secure Cross-Device Cloud Sync: Seamlessly sync model configs and skills across your devices via Chrome account; API Keys strictly physically isolated in local storage (never cloud-synced). 7. 👁️ Native Physical Control & Spatial Vision: Generates authentic user input events via Chrome DevTools Protocol (CDP); pairs with Set-of-Marks visual bounding boxes to conquer SPAs, nested Shadow DOMs, and dynamic iframes. 🔒 Sovereign Local-First Architecture (Zero-Cloud): - 100% Private (BYOK): Direct connection to your own standard model endpoints or local offline LLM instances. All API credentials and browsing data remain strictly inside your browser. No intermediary cloud servers, zero data tracking. - Credential Security: Physical secret isolation on local device; automatic token sanitization for skill variables; built-in financial domain blocklist; mandatory confirmation for sensitive actions. > 💡 Getting Started: This is a Bring-Your-Own-Key (BYOK) extension. Connect your own preferred model endpoint or local offline instance in Settings to get started.