Agent memory & persistence
What is Agent memory & persistence?
Researchers are building systems that let AI agents retain and recall information across multiple conversations, so they can remember context about users and tasks over time instead than starting fresh each time.
Persistent memory is essential for AI agents to become genuinely useful assistants rather than stateless tools, and the rapid open-source development suggests this capability is becoming a competitive baseline that will shape which agents people actually adopt.
References
- Excited to see the results of the agent memory leaderboard — Hacker News
- Mem0, Zep, or Letta: Your AI Agent Doesn't Have "Memory" — It Has One of Three Different Things — Medium: LLM
- Prinevo Memory: A Domain-Pluggable Context Graph for Long-Term Agent Memory — Hacker News
- PsychoAgent: An Affect-Sensitive Cognitive Architecture for Conflict-Aware Memory in LLM Agents — ArXiv
- TEPA: Revoking Stale Memories for Conflict-Robust Language Agents — ArXiv
- Total Recall: The Invisible Memory Layer for AI Agents — YouTube
- Why AI Agents Forget Everything (And How Mem0 Fixes It) — YouTube
- ABBEL: Natural-Language Belief States for Memory-Efficient Long-Horizon LLM Agents — YouTube