Agent Memory in Copilot Studio (Preview)

Copilot Studio agents can remember a person's preferences and context between conversations.

Each agent keeps a separate memory store per person, so one user's context is never visible to another
The agent captures preferences during a conversation, saves them as files in Microsoft-managed storage, and reads them back on later interactions
Makers switch it on per agent from the Build tab but cannot see what any user's memory holds
Users can ask the agent in chat what it remembers and tell it to forget something, or open a memory portal to view or delete everything at once
Memories are deleted automatically after 28 days without interaction, and memory is switched off in group chats and Teams channels
Turning memory off stops the agent using stored memories but does not delete them

Agents people use repeatedly, where re-explaining context every session is the friction

Your agent serves the same people week after week and should not need reminding who they are

Makers building conversational agents for an internal team

Staying informed is great. Staying ahead is better.

Pick a topic and leave with something you created.

Get new module drops and weekly AI strategy.