Agent Browser

CRHQ agents have access to a real web browser. They can navigate websites, interact with pages, fill forms, take screenshots, and extract information — all autonomously.

What Agents Can Do

• Navigate — Open URLs, go back/forward, reload pages
• Interact — Click buttons, fill inputs, select dropdowns, check boxes
• Read — Extract text, inspect elements, analyze page structure
• Screenshot — Capture full-page or viewport screenshots
• Automate — Complete multi-step web workflows (login, fill forms, submit)
• Test — Verify web applications, check layouts, validate functionality

How It Works

When an agent uses the browser:

1. A headless Chromium browser instance launches on your server
2. The agent navigates to the target URL
3. It takes a snapshot of the page (accessibility tree of interactive elements)
4. It interacts with elements by reference (click, type, select)
5. It re-snapshots after changes to see the updated state

Viewing the Browser

When an agent is using the browser, you can watch in real-time:

• A browser panel appears in the conversation interface
• Live screencasting via WebSocket shows what the agent sees
• You can observe the agent clicking, typing, and navigating

Use Cases

• Web scraping — Extract product data, prices, contact info
• Form automation — Fill out lengthy forms, submit applications
• Testing — Verify web applications across pages and flows
• Login and setup — Authenticate with services, configure settings
• Research — Browse documentation, compare products, gather information
• Screenshot evidence — Capture visual proof of issues or results

Authentication

Agents can log into websites using credentials stored in your connectors. Auth states can be saved so the agent doesn't need to re-login every time.