Memory Server for Claude, Vestige, Updated

Vestige is a local-first MCP memory server designed based on cognitive science, moving away from simple vector search methods. Through this v2.1.0 update, it now serves as a cognitive memory layer for Claude, going beyond a simple data store.

The key updates are as follows:

• Deep Reference: Through an 8-stage pipeline (hybrid search, reranking, spreading activation, FSRS trust scoring, temporal substitution, contradiction analysis, relationship evaluation, reasoning chain generation), it supports Claude in inferring relationships between memories and providing justification.
• Active Forgetting: Instead of indiscriminately storing all data, it maintains valid memories while intentionally managing outdated or unnecessary information to improve context efficiency.
• Self-check and Prediction: It has features that detect contradictions between data, predict what information will be needed next, and preemptively block incorrect answers.

This project is based on Rust and SQLite, and interacts with Claude through MCP (Model Context Protocol).