Your agent now remembers when it learned things—and asks before overwriting what it knows.
"When Did You Learn That?"
Your agent now tracks when it learned each piece of information. Ask questions like "when did you learn my preference for dark mode?" and get real answers instead of blank stares.
Memories include relative timestamps—"learned 3 days ago"—so you can understand the context of what your agent knows.
Smarter Conflict Handling
Previously, if you told your agent conflicting information, it would silently overwrite the old memory. Now your agent asks for clarification:
"I remember you preferred email updates, but you're now saying Slack. Which should I remember going forward?"
No more wondering why your agent "forgot" something you told it.
Memory Health at a Glance
A new status indicator on the Memory tab shows whether the memory service is connected:
- Green — Memory is working normally
- Red — Connection issue detected
If your agent seems forgetful, check the indicator first. You'll know immediately if there's a service issue rather than wondering what went wrong.
Better Tool Memory
Fixed an issue where memories weren't saved when your agent used tools without responding with text. Now all interactions are remembered, even when your agent is heads-down executing tasks.