MCP Memory Server

MCP Memory Server

Overview

MCP Memory Server is an open-source project currently in development that provides a memory storage, retrieval, and management system using Golang and Upstash's vector database. It offers a simple HTTP API for AI assistants to store and retrieve contextual information.

The project implements the MCP (Model Control Protocol) specification, allowing AI models to interact with external tools in a standardized way.

Features

Currently, the project is in development with three core API methods implemented:

1. Store Memory - Save information with automatic vector embeddings
2. Retrieve Memory - Semantic search for relevant information using natural language queries
3. Delete Memory - Remove specific memories by their ID

Technology Stack

• Backend: Go (Golang)
• Vector Database: currently supports Upstash Vector
• Transport: HTTP API

Getting Started

Installation

1. Clone the repository

   git clone https://github.com/bqx619/mcp-memory-server.git
   cd mcp-memory-server
   

2. Install dependencies

   go mod download
   

3. Create a .env file with the following variables:

   VECTOR_PROVIDER=upstash
   VECTOR_URL=your-upstash-vector-url
   VECTOR_TOKEN=your-upstash-vector-token
   HTTP_PORT=8080
   

Running the Server

go run .

The server will start on the configured port (default: 8080).

API Reference

Store Memory

Stores new information in the vector database.

Endpoint: POST /tools/store_memory Request Body:

{
  "content": "Memory content to store"
}

Response:

{
  "content": "Successfully stored memory with ID: abcd1234"
}

Retrieve Memory

Performs semantic search to find relevant information based on a query.

Endpoint: POST /tools/retrieve_memory Request Body:

{
  "query": "search term",
  "n_results": 5
}

Response:

{
  "content": "Found the following memories:
Memory 1:
Content: The first relevant memory
Id: abcd1234
Relevance Score: 0.92
Memory 2:
Content: Another related memory
Id: efgh5678
Relevance Score: 0.85
..."
}

Delete Memory

Removes a specific memory by its ID.

Endpoint: POST /tools/delete_memory Request Body:

{
  "memory_id": "abcd1234"
}

Response:

{
  "content": "Successfully deleted memory with ID: abcd1234"
}

Development Roadmap

• Adding authentication
• Implementing memory tagging and categorization
• Supporting additional vector database providers
• Enhancing search capabilities
• Adding memory expiration and TTL features
• ...

Contributing

As this project is currently in development, contributions are welcome. Please feel free to submit issues and pull requests.