nocturne_memory

Why we included this project

Most AI assistants forget what they knew the moment a conversation ends. This project gives them a durable memory server that persists across sessions, models, and tools. It speaks the MCP protocol, so any compatible client, from Claude Code and Cursor to Gemini CLI and OpenClaw, can read and write structured, graph-like memories rather than pushing everything into vector embeddings. The backing store is SQLite or PostgreSQL, which keeps things light and, more importantly, auditable. Every write is versioned and shown in a visual dashboard, so you can review diffs and roll back a change before it sticks. That makes it a good fit for teams that want reviewable agent memory instead of opaque retrieval, and for people who want one AI identity carried across several engines. It also works as a drop-in replacement for OpenClaw-style personal assistant backends, giving existing setups a straightforward migration path.