"Memory Core team on AI Agent Olympics Hackathon Hackathon"

Team Idea

The memory layer for multi-agent dev teams. Open-source MIT. SOTA on LongMemEval-S R@10 99.4 and LoCoMo R@10 88.8.

Team Members

Evanyuan-builder

evanyuanbuilder582

Submission

Memory Core: Role-Aware Memory for Agent Teams

An MIT-licensed role-aware memory layer that lets planner, coder, and reviewer agents share context through namespaces — not prompt-stuffing. SOTA: 99.4R@10 on LongMemEval-S.