How agents remember¶
Agents remember useful things across conversations — your preferences, ongoing projects, decisions you've made — and bring the relevant ones back when they matter. You don't have to repeat yourself every session.
What an agent remembers¶
As you work, an agent saves small, durable notes: facts ("prefers metric units"), context about a project, and conclusions you've reached together. These are scoped to that agent — each agent keeps its own memory, and memories never leak between organizations.
When you send a new message, the agent automatically recalls the notes most relevant to what you're asking, plus a running summary of what it knows about you. Older or less useful memories naturally fade over time, while important ones are reinforced each time they prove relevant.
Asking it to remember or forget¶
You can be explicit:
Remember that our fiscal year starts in April.
What do you remember about the Q3 launch?
Forget what I said about the vendor — that's no longer accurate.
The agent can remember, recall, list, update, and delete its memories on request. If you want a clean slate, ask it to forget a specific fact or clear out a topic.
Memory in read-only sessions¶
In a read-only session, the agent can read its existing memories but won't save new ones. It will tell you so plainly — something like "I can look this up, but I won't remember it for next time." Use a full session when you want the agent to retain what you discuss.
Privacy¶
- Memories belong to a single agent and are isolated to your organization.
- They follow your plan's retention window and are removed when they age out or when you delete the agent.
- Deleting an agent removes its memory.