Mnemon is a free MCP server tool developed by Mnemon Dev, designed to provide LLM-supervised persistent memory for AI agents. Built with Go, it runs on the mcp platform and ships as a single executable binary. Mnemon supports various distribution methods including Homebrew Cask for macOS, Go install for macOS, Linux, and Windows, building from source on macOS and Linux, and Docker container images. Its core function is to offer durable memory storage, enabling AI agents to retain context across sessions and interactions, enhancing their operational intelligence.

What does Mnemon connect to?

Mnemon serves as an MCP server, establishing connections with AI agent runtimes such as DeepSeek Harness, Claude Code, and OpenClaw, alongside any other compatible agent runtime. It functions as a tool provider via the MCP protocol, which standardizes how LLMs discover and invoke tools. Confirmed compatible AI clients include DeepSeek Harness, Claude Code, OpenClaw, Cursor, ZCode, Pi, and MiniMax Code. Mnemon utilizes SQLite as its underlying database for memory storage, offering persistent memory capabilities to these connected AI agents.

Core features and tools

Mnemon exposes several callable tools, providing LLM-supervised persistent memory. Its core primitives are remember, link, and recall. Additional tools identified include mnemon store create, mnemon log, mnemon status, mnemon search, mnemon import, and mnemon embed. Furthermore, it offers mnemon agency for functionalities related to durable agent work and peer collaboration, extending its capabilities beyond basic memory operations to support complex agent interactions.

Who is it for?

The primary audience for Mnemon is developers actively building AI agents. It provides these developers with essential LLM-supervised persistent memory capabilities, enabling their agents to maintain context, decisions, and facts across multiple interactions. This tool is particularly useful for those seeking to enhance agent intelligence through durable memory storage. This is not the right fit if you are not building AI agents.

How to get started

No credentials are required. Setup typically involves a single command, "mnemon setup", which defaults to a local, project-scoped installation.

Use cases

• Providing persistent, cross-session memory for AI agents, enabling them to retain context, decisions, and facts across multiple interactions and sessions.
• Facilitating the bulk import of historical chat logs or external context into a memory graph, which can then be semantically searched by AI agents.
• Supporting durable agent work and peer collaboration through its "mnemon agency" component, which adds project-local responsibility and effect admission to existing AI agents.

Mnemon offers AI agent developers a vital MCP server for LLM-supervised persistent memory, powered by SQLite. It stands out with its zero API keys model and a suite of tools for memory management and agent collaboration. The program extends AI agent capabilities by offering durable context retention and the ability to import external data into a searchable memory graph. Try importing historical chat logs into its memory graph for agent access.