MCP Memory Service

Example Use Cases

• 1

  Persistent or ephemeral semantic memory storage to save and later retrieve user interactions or contextual notes.

• 2

  Perform semantic similarity searches (nearest-neighbor retrieval) to find relevant past memories using transformer-based or dummy embeddings.

• 3

  Tag-based filtering and basic memory management (store, search by tag, delete) with configurable backends like in-memory for testing or ChromaDB for production.

Description

This repository implements a Model Context Protocol (MCP) Memory Service in Rust, exposing tools to store, retrieve, search by tag, and delete memories over stdio. It supports multiple storage backends (in-memory and ChromaDB) and multiple embedding models (dummy and ONNX transformer-based embeddings). The server is configurable via environment variables, uses asynchronous Tokio runtime, and includes a Node.js test script for integration testing.