MemStore, an AI memory platform developed by former Julep AI founders in India, offers a multi-layered approach to AI agent memory that prioritizes consistent outputs, flags contradictions rather than hiding them, and incorporates innovative data deletion techniques aligning with privacy norms.
- Agent memory structured in four layers for enhanced context
- Contradictions between facts are flagged, not auto-resolved
- Data deletion implemented by nullifying knowledge graph connections
What happened
MemStore’s co-founders, Diwank Singh Tomer and Ishita Jindal, detailed their approach to AI agent memory in an interview. They explained how the platform organizes memory across episodic, implicit, prospective, and semantic layers to provide agents with long-term memory capabilities beyond short-term context limitations.
They revealed that when conflicting information arises, instead of automatically resolving contradictions, MemStore flags them for awareness. Additionally, deletion requests for specific facts do not remove them from the knowledge graph physically but instead reduce the connection weights to zero, making the data effectively invisible in future queries.
Why it matters
Consistent and reliable memory in AI agents is critical as enterprises increasingly rely on AI collaboration across teams. MemStore’s system addresses the inherent variability in AI outputs by focusing on consistency rather than determinism, which mitigates risks of misinformation or repeated errors within organizational workflows.
The approach to contradictions and deletion aligns with regulatory demands for transparency and data privacy. By managing contradictions openly and enabling effective data forgetting mechanisms, MemStore supports compliance with data protection laws while enhancing trust in AI memory systems.
What to watch next
Monitoring how MemStore’s layered memory model performs in real-world enterprise deployments will be key, particularly its effectiveness in managing contradictions and enabling collaboration among AI agents across functional and team boundaries within organizations.
Future developments may focus on refining the deletion process for stronger guarantees and transparency, as well as exploring integration with India’s evolving regulatory landscape on data privacy and artificial intelligence ethics.