Skip to main content

Live Video Streaming

Granite streams the Windows desktop to your browser in real-time. See exactly what the agent sees.

How It Works

Technology: WebRTC for low-latency, high-quality streaming.

Video Panel Features

FeatureDescription
Live ViewCurrent desktop state
Full ScreenExpand for better visibility
ScreenshotCapture current frame
QualityAutomatic adjustment

Video Controls

Full Screen

Click to maximize the video

Screenshot

Capture the current view

Refresh

Reconnect if stream stalls

What You’ll See

During execution:
  • Desktop state - Applications, windows, dialogs
  • Mouse movements - Where the agent is clicking
  • Typing - Text being entered
  • Navigation - Window switches, scrolling

Latency

Video has minimal delay:
ConnectionTypical Latency
Same region100-300ms
Cross-region300-500ms
A slight delay is normal. HITL approvals aren’t affected since execution pauses.

Quality Settings

Video quality adjusts automatically based on:
  • Your internet speed
  • Server load
  • Network conditions
Higher quality = more bandwidth but clearer picture.

Troubleshooting

  • Check if WebRTC is allowed in your browser
  • Try refreshing the page
  • Disable VPN (may block WebRTC)
  • Try a different browser
  • Check your internet connection
  • Close other bandwidth-heavy applications
  • Try reducing video quality (if option available)
  • Refresh the stream
  • The driver may be initializing
  • Wait a few seconds
  • Check driver status in Driver Management
Audio is not streamed (video only). Agent actions don’t produce meaningful audio.

Browser Compatibility

BrowserSupport
ChromeFull
FirefoxFull
SafariFull
EdgeFull

Network Requirements

For reliable streaming:
  • Minimum: 1 Mbps download
  • Recommended: 5+ Mbps download
  • Firewall: Allow WebRTC (UDP)

Privacy

Video streams are:
  • Encrypted - TLS in transit
  • Not stored - Live only (recordings are separate)
  • Organization-scoped - Only your team can view

Tips

More screen space = easier to see details in the video.
When the agent navigates complex UIs, full screen helps.
The chat panel provides text descriptions. Use both for full context.

Next Steps