Piia Engram

This is a local-first memory layer that sits above your AI coding tools. Install it once, tell it your preferences and standards, and every MCP-compatible client (Claude Code, Cursor, Codex, Windsurf) reads from the same identity store. It exposes tools for AI to propose lessons, decisions, and context updates, but nothing becomes permanent until you approve it. All data lives in JSON and Markdown files under ~/.engram/ that you can inspect, edit, or version control. You'd reach for this if you're tired of re-explaining your code review bar every time you open a new chat window or switch between tools. The trust model is straightforward: no cloud, no vendor lock-in, and AI can only suggest changes to your memory, not write them directly.