Compartment
Provides encrypted, fully offline vector memory for AI agents, replacing their built-in memory with a secure and fast local alternative.
Compartment offers a superior, encrypted, and fully offline vector memory solution designed specifically for AI agents. It simplifies memory management with one-click installation and seamlessly takes over agents' built-in memory, ensuring all data is encrypted at rest, including embedding vectors. By operating entirely locally with a bundled embedding model, it guarantees absolute privacy and speed, never touching the network. Furthermore, it pre-seeds agents with thousands of self-contained reference facts, making them immediately useful without requiring prior training.
Key Features
01End-to-end authenticated encryption for all stored data, including embedding vectors
02Fully offline operation with an embedded embedding model and network runtime guard
03Advanced hybrid recall (vector + keyword + importance tiers) for precise memory retrieval
04Pre-seeded knowledge base with 6,718 reference facts on initialization
053 GitHub stars
06Automatic one-click installation and agent integration for various AI platforms
Use Cases
01Enhancing memory capabilities for Claude Code and Claude Desktop agents
02Integrating as a native memory provider for Hermes agents with zero setup
03Providing secure, local memory for OpenClaw and any other Model Context Protocol (MCP) client