home / mcp / mcp memory ts mcp server

MCP Memory TS MCP Server

Provides persistent, vector-based memory storage with semantic search for AI assistants via MCP.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "bobmatnyc-mcp-memory-ts": {
      "url": "http://localhost:3003/mcp",
      "headers": {
        "LOG_LEVEL": "info",
        "MCP_DEBUG": "0",
        "TURSO_URL": "libsql://your-database.turso.io",
        "OPENAI_API_KEY": "your-openai-api-key",
        "CLERK_SECRET_KEY": "sk_test_...",
        "GOOGLE_CLIENT_ID": "your-google-client-id",
        "TURSO_AUTH_TOKEN": "your-auth-token",
        "DEFAULT_USER_EMAIL": "[email protected]",
        "GOOGLE_REDIRECT_URI": "http://localhost:3002/api/auth/google/callback",
        "GOOGLE_CLIENT_SECRET": "your-google-client-secret",
        "ENABLE_EMBEDDING_MONITOR": "true",
        "EMBEDDING_MONITOR_INTERVAL": "60000",
        "NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY": "pk_test_..."
      }
    }
  }
}

You can run and interact with the MCP Memory Service to store, organize, and retrieve memory across multiple tenants, using vector embeddings for semantic search and a REST/JSON-RPC interface. This enables persistent memory for AI assistants, with secure multi-tenant access, easy web management, and remote HTTP access.

How to use

You use an MCP client to talk to the Memory Service over HTTP or via a local stdio channel. The service provides persistent memory storage with semantic search, multi-tenant isolation, and optional cloud integrations. When you connect, you can save memories, manage entities, search by meaning rather than exact keywords, and retrieve past interactions with context.

How to install

Prerequisites: ensure you have Node.js 18 or newer, a Turso-compatible database, and an OpenAI API key for embeddings.

Step-by-step installation and startup flow.

# Install dependencies
npm install

# Copy environment configuration (if you have a template locally)
cp .env.local .env

# Build the project
npm run build

# Start development server
npm run dev

Additional sections

Configuration and usage notes help you tailor the MCP Memory Service to your environment. The setup supports both local development and remote HTTP access with Clerk authentication for multi-tenant scenarios.

Remote HTTP access with OAuth enables web applications to instance the MCP server securely via tokens. You can start the remote server and issue MCP requests over HTTP, then manage memory, entities, and searches from your applications.

Web interface and REST API are provided for visual memory management and programmatic access. The web UI runs on a separate port and offers authenticated access, visual search, and entity relationships.

Environment variables control core behavior, embedding generation, scheduling of background tasks, and cloud integrations. Ensure you provide appropriate values for database connection, embeddings, and authentication keys.

MCP server configuration sample

{
  "mcpServers": {
    "mcp-memory-ts": {
      "type": "http",
      "url": "http://localhost:3003/mcp",
      "args": []
    }
  }
}

Available tools

[]

No explicit MCP tools described in this guide; features are described as capabilities rather than discrete tools