HeurChain

HeurChain is a memory system for AI agents that lets them maintain structured memories across different sessions and models. Instead of losing context between runs, an agent backed by HeurChain can recall important information over the long term. It organizes memory in three tiers and applies ACT-R cognitive decay principles, so frequently used information is retained while unused data gradually fades — a design that keeps the memory store relevant without letting it grow without limit. For retrieval, HeurChain combines BM25 keyword search with vector search, pairing exact-match and semantic lookup to return relevant memories with fast access times. It's distributed as an npm package, which makes it straightforward to add to JavaScript or Node projects, but it's infrastructure aimed at developers rather than a tool with an end-user interface. Wiring it into an existing agent stack takes integration work, and tuning the decay behavior may need some experimentation to match your retention needs. HeurChain suits engineers building agents, assistants, or long-running AI workflows that need durable, well-organized memory.

Developers use HeurChain to give AI agents persistent, tiered memory that recalls important context across sessions and models while decaying unused data.