Tinycortex

TinyCortex is a local first AI memory engine shipped as an open source Rust crate. It gives AI agents a long term memory that behaves more like a brain than a vector store: low value memories decay over time, while knowledge that is recalled and interacted with is reinforced and rises to the top. The goal is to stop agents from drowning in stale context, which degrades responses and inflates costs. The engine ingests content, canonicalizes and chunks it, scores what is worth keeping, and compresses it into a hierarchical summary tree. Retrieval combines vector, keyword, graph and tree search to return a focused, explainable slice of history. The documentation also covers a job queue, a diff layer, entity and graph extraction, source connectors, goals and tool memory, and conversation handling. Memory is stored locally as inspectable files, and signals such as views, replies and reactions inform what the engine keeps. It is aimed at developers building agents in Rust who want memory under their own control. Installation is a single cargo add tinycortex, and the docs include a quickstart with an in memory store, insert and search calls. The documentation separates features of the open source crate from those of a hosted TinyCortex platform, which offers a managed API and language SDKs and is in closed alpha with access by request. The crate is open source. The documentation site publishes no pricing for the hosted platform.