Hippocampus – self-hosted memory shared by every Claude Code session (github.com)

Hippocampus has been introduced as a self-hosted memory server designed for Claude Code and other MCP clients, enabling persistent long-term memory across multiple projects and sessions. Unlike previous setups where memory was limited to individual project folders, Hippocampus centralizes memory storage as plain Markdown files indexed with SQLite. This allows for a hybrid search function that combines BM25 keyword matching, vector embeddings, and a knowledge graph to retrieve relevant information efficiently. It enhances productivity by allowing Claude Code to recall and integrate learned information seamlessly, reducing repetitive explanations and improving context retention. This development is significant for the AI/ML community as it empowers developers to manage and utilize their AI models more effectively, fostering enhanced collaboration across different environments and machines. The self-hosted nature of Hippocampus, which runs locally or on a VPS, ensures user privacy by storing all data on personal servers without relying on third-party APIs. Key features include automatic memory recall on each prompt, status tracking of open tasks, and the ability to curate memories at the end of sessions, ultimately streamlining the development process and improving workflow efficiency in AI projects.