Geniuz
Tempreon
Claude by Anthropic
ChainMemory
Codex by OpenAI
Cursor
Gemini
MemMachine
Nullstack
Deno
Geniuz is an open-source MCP server that gives Claude, Hermes, and other AI agents a memory that lasts. Normally every conversation starts from zero โ you re-teach your context, preferences, and decisions each time. Geniuz keeps them: the standards you set, the calls you made, and the reasons behind them, recalled by meaning across sessions and compactions. It runs entirely on your own machine โ semantic search via a bundled ONNX model, no cloud, no account, no API keys โ and plugs into Claude Desktop through Anthropic's MCP standard on macOS, Windows, and Linux. Free and MIT-licensed. Built by mVara, who run their own fleet of AI agents on it every day.
geniuz mcp install wires it into Claude Desktop instantly โ zero config
Geniuz NullstackNo Geniuz videos yet. You could help us improve this page by suggesting one.
Geniuz's answer
Geniuz's answer
Built in Rust. On-device semantic search runs on ONNX Runtime with a bundled embedding model and tokenizer, so inference happens locally with no network calls. Memory is stored in an embedded SQLite database. It connects to AI clients through Anthropic's Model Context Protocol (MCP), and ships with a terminal UI (ratatui) plus native desktop apps for macOS, Windows, and Linux.
Geniuz's answer
Geniuz came out of mVara's own work running a fleet of AI agents. The recurring problem was that every agent forgot everything between sessions โ every morning was Day 1, re-teaching context, standards, and decisions. mVara built Geniuz to solve that for themselves: a local memory layer their agents share and hand off context through, every day. The product is what the company actually uses, which is why it's shaped around real daily use rather than a demo.
Geniuz's answer
Developers, AI power users, and teams who work with Claude every day โ people running Claude Desktop, Claude Code, Cursor, Windsurf, or their own agent frameworks who are tired of re-explaining their context each session. It's especially suited to fleet and agent operators coordinating multiple AI agents that need shared, persistent memory, and to anyone privacy-conscious who wants that memory to stay on their own machine.
Geniuz's answer
Most memory tools are hosted services or developer SDKs you wire into code. Geniuz is local-first and zero-config: install it, run one command, and Claude Desktop has memory โ no account to create, no keys to manage, no data sent anywhere. It's open source under the MIT license and fully functional for free. If you value privacy, want to own your data, and want memory that works in the app you already use rather than a framework you have to build around, Geniuz is the simplest path.
Geniuz's answer
Geniuz runs entirely on your own machine. Memory is recalled by meaning, not keywords, through a semantic search model bundled right into the app โ no cloud, no account, no API keys, and nothing leaves your computer. It plugs into Claude Desktop through Anthropic's MCP standard with a single command, exposing just three tools: remember, recall, and recent. Memories survive session boundaries and context compaction, so your AI keeps the standards you set and the reasons behind your decisions instead of starting from zero every conversation.
Tempreon - A personal memory layer for your AI tools, connected over MCP.
Deno - A secure runtime for JavaScript and TypeScript built with V8, Rust, and Tokio.
Claude by Anthropic - A family of foundational AI models
ChainMemory - Portable, verifiable memory for AI agents โ works across ChatGPT, Claude, Gemini and any MCP client
Codex by OpenAI - AI that writes the code for you
Cursor - The AI-first Code Editor. Build software faster in an editor designed for pair-programming with AI.