simplemem-skill

SimpleMem provides a robust, persistent memory layer for LLM-powered agents, enabling them to maintain context across disparate sessions. By implementing semantic lossless compression, it transforms raw dialogue into atomic, self-contained facts, allowing agents to recall specific decisions, commitments, and complex information uncovered during long-term interactions. It is designed for developers, researchers, and power users who require AI assistants to evolve alongside their projects rather than starting from a blank slate in every chat.

The system handles coreference resolution, temporal anchoring, and multi-tenant isolation, ensuring that stored data remains relevant, searchable, and secure. Whether you are using Claude Desktop, Cursor, or an MCP-compatible environment, SimpleMem functions as an intelligent repository that proactively saves valuable dialogue and retrieves context before you even ask. It supports hybrid retrieval methods, combining semantic search with keyword matching to deliver highly accurate, noise-free information.

• Advanced semantic compression and coreference resolution to replace pronouns with named entities.

• Support for multi-session and cross-conversation context preservation with persistent storage in LanceDB or SQLite.

• Multimodal memory capabilities, enabling the storage and retrieval of text, images, audio, and video experiences.

• High-performance architectural design, significantly outperforming legacy memory solutions on the LoCoMo benchmark.

• Seamless MCP (Model Context Protocol) integration for production-ready, streamable HTTP transport.

• Flexible CLI tools for manual memory management, including stats, clearing data, and bulk JSONL imports.

• Utilize the add command with ISO 8601 timestamps to preserve the chronological history of complex decisions.

• Leverage the query command with optional reflection to enable the agent to analyze retrieved information for deeper insights.

• Configure OpenRouter API keys in src/config.py to enable high-quality embedding models and LLM processing.

• Use custom table names to partition memory contexts for different project domains or research threads.

• Regularly monitor memory statistics using the stats utility to manage storage growth.

• Ensure all dependencies are met via the provided requirements.txt to maintain compatibility with your local environment.