Show HN: Agam – Activation-based memory for Claude Code, not retrieval (github.com)

Agam has been introduced as a powerful tool designed to enhance Claude Code sessions by integrating an activation-based memory system. This innovative layer automatically injects relevant contextual information—such as entities related to projects, services, and past decisions—during interactions via a UserPromptSubmit hook. By using a knowledge graph, Agam allows Claude Code to respond based on historical context rather than requiring file searches for every query. The user simply seeds the graph with existing session transcripts, which Agam then maintains in the background, promoting a more efficient coding experience. The introduction of Agam is significant for the AI/ML community as it streamlines workflows, particularly for developers working with Claude Code. It minimizes repetitive context retrieval while optimizing session performance. Key technical features include a SQLite-based knowledge graph that supports full-text search, hooks for real-time context injection, and a streamlined installation process suited for macOS users. With a focus on keeping continuous context without needing an external API key, Agam exemplifies a move towards more intuitive interactions with language models, ultimately facilitating a richer and more personalized development environment.