Browser Agent: Privacy Vision 作成者: Chandan Meher

Browser Agent is a privacy-preserving visual AI assistant that helps automate web interactions, form filling, and web navigation without exposing your personal information.


Unlike conventional AI web agents that send full page content, screenshots, and form data directly to cloud LLMs, Browser Agent guarantees:
- Local Vault Storage: All sensitive credentials and personal data (names, emails, phone numbers, addresses) are stored locally in your browser's encrypted IndexedDB using AES-GCM-256.
- Client-Side Redaction: Sensitive data fields are automatically replaced with synthetic tokens (<NAME_1>, <EMAIL_1>) on your device before any prompt is dispatched to the backend.
- Secure Substitution: Values are decrypted and injected back into web pages locally at execution time.

• Visual & Semantic DOM Understanding: Detects interactive buttons, inputs, dropdowns, and checkboxes.
• Natural Language Execution: Tell the agent to "Fill the registration form" or "Navigate to settings" and watch it execute the steps.
• Human-in-the-Loop Safeguards: Highlights target elements on screen so you can inspect actions as they happen.

1. Click the Browser Agent icon in your Firefox toolbar.
2. Under the Vault tab, verify or customize your profile information.
3. Open any webpage with a form or task.
4. Open the extension popup, enter your instruction in the Agent tab, and click Run Agent.