engrim
The Universal Cross-Model Episodic Memory Standard. Local-first, project-scoped SQLite engine for Google Antigravity, Claude Code, Cursor, Codex CLI, GitHub Cop
What is engrim?
The Universal Cross-Model Episodic Memory Standard. Local-first, project-scoped SQLite engine for Google Antigravity, Claude Code, Cursor, Codex CLI, GitHub Copilot CLI, and OpenCode. Zero cloud lock-in.
Overview of engrim
What it is
Engrim is a local-first, project-scoped SQLite memory engine designed for autonomous AI coding agents and heavy software development. It provides a universal cross-model and cross-agent episodic memory standard, allowing developers to switch between models and environments (Google Antigravity, Claude Code, Cursor, Codex CLI, GitHub Copilot CLI, OpenCode, and Windsurf) on the same codebase without losing architectural decisions, user constraints, or project state. It addresses the “200,000-token context window trap” by replacing raw transcript replay with approximately 4,000 characters of high-precision, curated episodic working memory stored in a single local SQLite file (~/.engrim/memory.db).
Key features
- Local-first SQLite storage with zero cloud lock-in.
- Hybrid retrieval combining BM25 lexical search and model2vec vector search, returning results in ~30ms.
- Automatic detection and setup for multiple agent environments via
engrim setup. - Diagnostic health check and self-healing via
engrim doctorwith--fixand--jsonoptions. - MCP server support for tool integration (
engrim serve --mcp). - Lifecycle hooks (boot, capture, recall, continue-as-clear) for seamless memory injection and persistence.
- Provenance tracking of origin agent for each memory record.
Use cases
- Maintaining persistent project memory across multiple AI coding agents and sessions.
- Switching between different models (e.g., Claude, Gemini, Codex) without losing context or prior decisions.
- Reducing token costs and attention dilution by replacing full transcript replay with compact episodic summaries.
- Enabling collaborative development where multiple agents share a common memory store.
- Integrating with CI pipelines for health monitoring via JSON diagnostic output.
Getting started
bash pip install engrim # Auto-detect and configure all installed agent environments engrim setup # Run health diagnostic engrim doctor For explicit platform setup, see the repository README.
License and pricing
Licensed under MIT. Open source and free to use.
Based on the project's public README and repository metadata.