Skip to main content
Javier - Developer Relations
Community Manager
September 1, 2026
Sticky

AI Agents Feature Roundup #2: What you can test today - Aug 31st

  • September 1, 2026
  • 0 replies
  • 13 views

 

Last month was about getting your agent into Front as a teammate. This month is about running it with confidence. Everything below shipped in late August and is available to test now.

 

See every reply your agent has touched with Reply Hub

The Replies Hub gives you one place to review what your AI teammate drafted and sent, along with quality signals to help you spot problems at the reply level. If you have been piecing this together from individual conversations, this view saves you time and consolidates where you find this information.
 

Learn more

 

Tell your agent when it got something wrong

Your team can now give feedback directly on AI teammate drafts and auto-sent replies from the inbox. No separate tool, no spreadsheet of complaints. The people closest to the work flag what is off the moment they see it.

 

Jump straight from a reply to the full story

Drafted and auto-sent replies now show a clear disclosure in the inbox and deep-link into the Replies Hub. One click from "why did it say that" to the answer.
 

 

Hit pause without tearing anything down

Pause an AI teammate and it stops taking action immediately. Useful when you are shipping a change, investigating something odd, or just want a quiet weekend. Unpause when you are ready, and all your settings, permissions, and credentials for the agent remain the same.


 

 

Give your agent a face

AI teammates now support rich animated avatars, including gaze tracking that follows your cursor, plus an avatar editor to make it your own. These small tweaks make an agent feel like part of the team rather than a process running in the background.

 

What's next

Agent testing against simulated conversations, AI agent reporting, and diagnosis support are all in development. Testing in particular is the one we are most eager to get into your hands, since validating a change before deploying it is still the hardest part of running an agent.
 

If you have questions or feedback on any of the above, please create a post to let us know!

For previously released features, please refer to: