PMB (Personal Memory Brain)

PMB is a local-first persistent memory solution designed for AI coding agents such as Claude Code, Cursor, and Codex. By operating entirely on local machines, PMB eliminates the need for cloud interactions, API keys, or executing language models in critical operational paths.

Key Features:

• High-Performance Recall: Achieves a 94.5% retrieval recall at rank 10 using the LoCoMo benchmark along with an average response latency of 70ms.
• Rich Multilingual Support: Out-of-the-box capability to handle 50+ languages and a robust multi-language embedder for effective extraction and recall across different languages.
• Seamless Integration: Supports integration across multiple AI agents through the Model Context Protocol (MCP), enabling a shared workspace for memory.
• Offline and Private: Operates without network dependencies, ensuring data privacy and control since all memory is stored on the local disk.

Benefits:

PMB addresses the common issue of AI agents forgetting context between sessions. This project provides an elegant solution by recording critical information and decisions, enabling AI agents to maintain memory across different tools and machine reboots. Users can command their agents to remember or retrieve stored facts, improving workflow efficiency.

Installation and Usage:

1. Setup: Install with a straightforward command: pip install pmb-ai followed by configuring to connect your AI agent.