engrim

engrim: The Universal Cross-Model and Cross-Agent Episodic Memory Standard

engrim is a local-first, project-scoped SQLite memory engine designed for seamless integration across diverse AI models and environments, specifically optimized for Google Antigravity, Claude Code, Cursor MCP, and Windsurf. This innovative memory store empowers developers to maintain their architectural decisions, user constraints, and project states without dependency on a single AI vendor or proprietary cloud solutions.

Core Value Proposition

With engrim, avoid the pitfalls of attention dilution that come with lengthy context windows. This tool provides 4,000 characters of curated episodic working memory, which facilitates:

• Versatile Agent Switching: Effortlessly switch between AI models mid-project without losing progress. For example, transition from Gemini 3.8 in Antigravity to Claude 3.7 Sonnet in Claude Code, ensuring your agent retains all relevant context.
• Memory Management: Externalize decisions, constraints, and project state easily. Clear agent sessions when needed and reload context seamlessly.
• Efficient Context Management: Utilize advanced SQLite features combined with static vector embeddings for optimal performance in memory retrieval.

Empirical Proof

In extensive testing across 105 sessions on a 50,000-line algorithmic trading codebase, engrim proved its reliability:

• Recorded 153,000 tokens of work consolidated into an efficient memory pack under 1,000 tokens.
• Achieved a 99%+ reduction in context reload costs during session restarts while ensuring zero regressions across 186 unit tests and no context loss when switching between models.

Architecture Overview

The architecture of engrim supports diverse agent environments, integrating hooks and adapters for:

• Google Antigravity
• Claude Code
• Cursor and Windsurf

Graphical representation of the core components demonstrates these integrations in action.

Features and Commands

engrim provides a range of commands that facilitate interaction and memory management:

• Add Memory: engrim add -t decision -s "..."
• Recall Memory: engrim recall -q "database"
• Context Management: engrim context [-b 4000]
• Agent Setup: engrim setup --all to configure all supported environments with ease.

Incorporate engrim into the workflow to manage memories effectively and maintain clear communication between multiple AI agents. Capture significant decisions during development and leverage features like session verification to ensure all crucial data is retained.

Security and Privacy

• Local and Offline: All memories are stored securely in a local SQLite database without the risk of cloud exposure.
• Data Protection: Implement POSIX file permissions for enhanced security and prevent accidental commits in version control systems.

engrim stands as a comprehensive solution designed to enhance productivity and memory management across AI-driven coding environments, ensuring efficiency, security, and enhanced project continuity.