Web Navigator is an LLM-driven browser that allows AI models to browse the web directly within chat sessions. The AI can open pages, navigate URLs, read content, fill forms, take screenshots, and execute JavaScript — all without leaving the conversation.
How to Use Web Navigator
- In the chat prompt area, click the + button.
- Select Navigator from the menu (monitor icon).
- A Navigator badge appears in the prompt area.
- Type your request (e.g., “Go to example.com and find the pricing information”).
- The AI opens a browser session and interacts with the web page.
What You Can Ask the AI to Do
With Navigator active, the AI can:
- Open websites: “Visit example.com”.
- Navigate to specific pages: “Go to the pricing page”.
- Read page content: “Get me the main article text”.
- Fill forms: “Enter my email and submit the form”.
- Click buttons: “Click the Sign Up button”.
- Take screenshots: “Show me what’s on screen”.
- Run JavaScript: “Execute this code on the page”.
- Scroll and explore: “Scroll down to see more content”.
- Extract data: “Get all the product names and prices from this table”.
Session Management
- Each Navigator request creates a browser session.
- Sessions persist across multiple prompts within the same conversation.
- Maximum 5 concurrent sessions to manage resources.
- Sessions auto-close after inactivity (TTL-based cleanup).
- Each session has its own isolated browsing context (cookies, storage).
Visual Feedback
When the AI uses Navigator, tool results show:
- Action description: What the AI did (navigate, click, get_html, etc.).
- Current URL: The page the browser is on.
- Screenshot preview (expandable): Visual capture of the page.
- Loading indicator: Shows while the tool executes.
Deactivating Navigator
Click the X on the Navigator badge in the prompt area to deactivate it for subsequent messages.