Show HN: Agent Recall – Open-source, local memory for AI agents (SQLite/MCP) (github.com)

A new open-source project called Agent Recall has been announced, providing persistent memory capabilities for AI agents in local environments using SQLite. Traditionally, AI agents lose contextual information between sessions, making them forgetful about key details such as user preferences and project statuses. Agent Recall addresses this by allowing agents to remember important facts and context across sessions, dramatically improving their efficiency. The project is rooted in production use, having been developed to fix issues encountered while operating more than 30 concurrent AI agents at a digital agency. Agent Recall distinguishes itself from existing memory solutions through its unique scope hierarchy, enabling agents to manage different roles for the same individual across various projects. With features like proactive saving of context during interactions, bitemporal memory archiving, and local storage with minimal dependencies, it enhances the usability of AI agents significantly. Furthermore, it integrates AI briefings through LLM summarization, ensuring agents start each session fully informed. This development is particularly significant as it opens avenues for more sophisticated multi-agent systems while keeping data secure and manageable, setting a new standard for memory management in AI applications.