MemCurrent: Designing Persistent Memory for Production Agents

Summary

Enterprise agents operating at scale must balance answer quality against cost and latency, a trade-off that memory can address if designed carefully. MemCurrent is a persistent memory architecture that stores both episodic memory (raw session traces) and derived memory (distilled learnings in a structured file system indexed by memory.md). It accesses memory through three patterns: plugins for automatic context logging and injection at session start, tools the agent calls explicitly during a session, and an offline dreaming pipeline that consolidates memory asynchronously. On an internal benchmark built from 114 real-world test cases across 29 users, MemCurrent triples the agent’s pass rate compared to no memory while reducing token consumption by 31% and latency by 26%. It also matches full-context accuracy on open-source benchmarks at roughly 12x fewer input tokens, placing it on the cost–accuracy Pareto frontier.

Takeaways

1. 1

   MemCurrent uses three memory access patterns: plugins for automatic logging and injection, tools for agent-driven reads and writes, and an offline dreaming pipeline for background consolidation. Each balances immediacy of updates against runtime overhead.

2. 2

   On an internal benchmark built from 114 real enterprise test cases across 29 users, MemCurrent triples the agent’s pass rate over no memory while reducing token usage by 31% and latency by 26%.

3. 3

   Derived memory drives most of the improvement; combining online tools with the offline dreaming pipeline produces the best extraction strategy, as explicit requests take effect immediately and dreaming handles implicit distillation.

FAQ