AI Memory Bible — Local Memory for Claude Desktop, Cursor & Claude Code (MCP)
Tired of re-explaining your project to your AI every new chat?
AI Memory Bible gives Claude Desktop, Cursor and Claude Code a long-term memory that lives only on your computer — one SQLite file. No cloud account, no API key, no subscription, no network calls.
What's inside
- Local MCP memory server with 10 tools: remember, recall, search, update, forget, list, projects, briefing, export, import
- Fast ranked full-text search (SQLite FTS5 / BM25), projects and tags
- One-command installers for Windows (PowerShell) and macOS/Linux (bash) that register the server in Cursor and Claude Desktop and back up your config first
- memory-bible doctor self-check that verifies the install and database and prints the exact config for your PC
- Setup guide PDF (English + Korean quick start) with exact config file paths per OS and a troubleshooting FAQ
- Templates: profile, projects, decisions, and a "memory rules" prompt to paste into your AI's instructions
- JSON export/import for backups and moving to a new PC
Requirements: Python 3.9+ (free), and Claude Desktop, Cursor or Claude Code. Windows 10/11, macOS, or Linux.
Honest testing note: Automated tests, a clean install and MCP-protocol checks (official MCP Python SDK client and MCP Inspector) pass on Linux. The comparison image is a real scripted session: same question in a new chat without vs with the memory server; the assistant replies in it were typed by the tester acting as the AI, using only what each session contained, and tool outputs are raw. The Windows/macOS installers were linted (and the PowerShell one run with PowerShell 7 on Linux) but not yet run on real Windows/Mac machines. ChatGPT is not supported directly (use export + Custom Instructions).
License: commercial single-user license (use on all your own computers; no redistribution).
Support & refunds: Support is via the documentation only; if it doesn't work after following the guide, you get a refund.
Made by Jipbap Kitchen. Free mini version (Memory Lite, MIT) available on GitHub — https://github.com/jipbapkitchen/memory-lite-mcp .