Shared long-term memory for AI agents can cut repeated setup
Personal become less useful when every new session requires the same , project details, and ps to be entered again. That repeated setup weakens and can waste tokens. with memory can work for light use, but they may lose usefulness when the is reached.
Note tools such as Mem can store information, but they do not always help an agent act on that information. Open Campus uses an agent setup where several small agents share one instead of each keeping separate memories, so a resume agent and a planning agent can both use the same user history. The setup is described as being built on the Animoca Minds framework.
Shared memory also creates a hard problem: when two agents hold different versions of the same fact, there may be no clear way to decide which one is correct.
Key points
- Re-entering the same in every session wastes effort and tokens.
- Custom GPT memory can help with light tasks but may struggle around the .
- Note tools can store information but may not make agents use it effectively.
- A shared lets multiple small agents use the same user history.
- Memory conflicts need a clear way to choose or update the correct fact.