engrim

Engrim is a universal, local-first SQLite memory engine specifically engineered for autonomous AI coding agents, designed to solve the '200,000-token context window trap'. It provides high-precision, curated episodic working memory, allowing developers to switch between various models and environments on the same codebase without losing architectural decisions or project state.

engrim sits in the Memory & Context category, so it is usually better judged by workflow fit, deployment style, and team needs than by isolated features alone. Based on the available metadata, it leans most clearly toward Precision Working Memory: Replaces raw transcript replay with ~4,000 characters of high-precision, curated episodic memory, injected at boot with zero attention dilution.. A representative use case is Maintaining Project Context Across Agent Sessions: Developers can use Engrim to prevent LLMs from losing architectural decisions, user constraints, or project state when chat sessions are cleared or models are switched.. The listed license is MIT. GitHub stars are currently 299. Relevant tags include Episodic Memory, LLM Context Management.