Consolidation Memory

关于

Consolidation Memory provides a robust, local-first memory system designed to prevent AI agents from losing context between sessions. It captures 'episodes' (facts, solutions, preferences) and intelligently synthesizes them into structured knowledge documents using a background LLM. This allows AI agents to recall relevant past interactions, adapt to user preferences, and offer more informed responses, all while keeping user data securely on their machine with no subscriptions required.

主要功能

• Local-first persistent storage using SQLite and FAISS
• Automated knowledge consolidation via background LLM clustering and synthesis
• Semantic recall with priority scoring (surprise, recency, access frequency)
• MCP Server integration providing six tools for Claude Desktop/Code/Cursor
• Support for multiple embedding and LLM backends (FastEmbed, LM Studio, Ollama, OpenAI)
• 1 GitHub stars

使用案例

• Providing long-term memory and context for AI agents in development environments
• Enhancing Claude Desktop, Claude Code, and Cursor with persistent knowledge recall
• Storing and consolidating technical solutions, project specifics, and user preferences for AI assistance