mnemon plugin: LLM-Supervised Persistent Memory for AI Agents

Graph-based persistent memory for CLI agents, with LLM-supervised recall across sessions and a single-binary deployment.

What this plugin does

The mnemon plugin provides LLM-supervised persistent memory for AI agents, enabling graph-based recall and cross-session knowledge. It integrates with DeepSeek Harness and other CLI agents, offering a single binary for memory management.

Mnemon is a persistent memory system for AI agents, supervised by the LLM itself. It uses three primitives—remember, link, recall—to form an intent-native protocol. The system stores knowledge in a graph structure, allowing for cross-session recall and compound interest effects. It integrates with DeepSeek Harness through the dsh-mnemon plugin, layering runtime memory and managed documents into a three-tier memory system. The mnemon plugin is designed for CLI agents and supports multiple integrations.

Documented capabilities

Graph-based memory

Mnemon builds a knowledge graph with insights and edges across multiple graph types, enabling structured recall and relationship discovery.

LLM-supervised lifecycle

Hooks drive the memory lifecycle, while the host LLM makes judgment calls and the binary handles deterministic computation.

Cross-session knowledge

Memory persists across sessions, allowing agents to accumulate knowledge over time and use it in future interactions.

Practical use cases

Persistent agent memory

Use the mnemon plugin to give your AI agent long-term memory, so it remembers past interactions and user preferences.

Knowledge graph building

Leverage mnemon's graph-based recall to build and query a knowledge graph of insights and relationships from your data.

Multi-agent integration

Integrate mnemon with various CLI agents like Claude Code, OpenClaw, and DeepSeek Harness for consistent memory across tools.

Review before you install

Read the source, recent releases, requested permissions, network access, and maintainer history before trusting a third-party plugin.

Frequently asked questions

How does the mnemon plugin work with DeepSeek Harness?

The mnemon integrates with DeepSeek Harness through the dsh-mnemon, layering runtime memory, managed documents, and long-term memory spaces into a supervised three-tier memory system.

What are the main features of the mnemon?

The mnemon offers graph-based recall, cross-session knowledge, and LLM-supervised lifecycle management, all in a single binary.

Which agents are compatible with the mnemon?

The mnemon works with Claude Code, OpenClaw, Codex, Cursor, and DeepSeek Harness, among other CLI agents.