Cursor Memory

You can store, organize, and recall memories for the Cursor IDE using this MCP server. It provides persistent memory that your AI assistant can access across sessions, with support for global or project-specific memories, tag-based organization, and efficient CRUD operations via a lightweight SQLite backend. Docker deployment makes setup simple and repeatable.

How to use

You interact with the Memory MCP server through an MCP client integrated into Cursor. Start by launching the server, then configure your client to connect to it. Once connected, you can create memories with optional scope and tags, retrieve memories on session start, search by keywords or tags, and update or delete memories by ID or by their description.

How to install

Prerequisites You need Docker and Docker Compose installed on your machine.

1. Deploy the Memory MCP server with Docker Compose.

docker-compose up -d --build

Configuration and startup notes

Configure the MCP client to connect to the Memory MCP server. The server exposes an SSE endpoint at port 8081 by default.

Example client configuration (HTTP MCP connection)

{
  "mcpServers": {
    "cursor-memory": {
      "url": "http://localhost:8081/sse",
      "transport": "sse"
    }
  }
}

Additional notes

Environment variables you may configure for runtime behavior include the HTTP server port, data directory for the SQLite database, and the environment mode. Adjust these if your deployment requires non-default values.

Architecture

The Cursor IDE communicates with the Memory MCP Server over SSE. The server is built with Express and SQLite, persisting memories in a local SQLite database.

Port configuration

The Memory MCP Server runs on port 8081 by default to avoid conflicts with other MCP servers.

Development options

For local development without Docker, you can install dependencies and run in development mode. The project uses Node.js and npm for tooling.

Available tools

memory_store

Store a new memory with optional scope and tags.

memory_recall

Retrieve all memories (typically called on session start).

memory_search

Search memories by keyword or tag.

memory_list

List all memories with index numbers for easy reference.

memory_update

Update a memory by ID or by its context.

memory_delete

Delete a memory by ID or context.