MUSE gives AI coding agents (Claude Code, OpenClaw, Cursor, Windsurf, Gemini CLI, Codex) persistent cross-session memory and multi-role governance — all through plain Markdown files.
Cost / License
- Free
- Open Source (MIT)
Platforms
- Mac
- Linux
Taiwan

Apps similar to SAME (Stateless Agent Memory Engine) include Tolaria, which is free and open source. Other options are Model Context Protocol, Claude-Mem, Qarinah and LedgerMind . You can choose from more than 10 Large Language Model (LLM) Tools like SAME (Stateless Agent Memory Engine) for the web, Mac, Windows, Linux and self-hosting.
MUSE gives AI coding agents (Claude Code, OpenClaw, Cursor, Windsurf, Gemini CLI, Codex) persistent cross-session memory and multi-role governance — all through plain Markdown files.

Every AI agent forgets the moment a session ends, and you become its memory: re-explaining the project, the conventions, what was already tried. brethof-brain is the memory instead.



The knowledge base for ChatGPT and Claude. Save notes once, connect via MCP, stop re-explaining context every conversation.




Haki gives any AI agent a memory that lasts for months dated, sourced, and provable. Self-hosted on your own Postgres, Apache-2.0, no amputated edition.

Tracebrain turns your conversation into your memory: what you're working on, what you promised, and everyone who matters. In a vault of your own. Usable by any AI you choose.

MemoryPlugin gives AI the ability to remember you — your preferences, history, and context — across chats and tools. No more starting from zero. Enjoy smarter, faster, and more useful conversations that help you accomplish more.



An AI assistant that permanently stores user corrections and preferences, applying them across every future session and project to eliminate repeated teaching, personalize outputs for creators, students, developers, teams, and maintain reliable consistency.
Maximem Synap is a managed long-term memory layer for AI agents. It stores what an agent learns across sessions and returns the relevant parts at inference time, instead of re-sending an entire conversation history into the context window.