Deployment Guide

Choose your deployment mode and get started in minutes.

Deployment Modes

Choose the deployment that fits your privacy and operational needs. All modes support ONE LLM Mode — a toggle that lets your MCP client's own LLM handle memory extraction, eliminating extra LLM API calls.

Cloud

Standard

Fully managed on Neuromem Cloud. Just add your API key to any MCP client — zero setup, instant access.

ClientCloud

Use case: Quick integration, zero ops

Self-Hosted

Maximum

Run entirely on your machine with Docker. Data never leaves localhost. Supports ONE LLM Mode — no extra LLM API needed.

ClientMCP ServerLocal PG

Use case: Privacy-sensitive, personal dev

Hybrid Encrypted

High

Local encryption + cloud API. No database needed — just an API key and your encryption key. Supports ONE LLM Mode.

ClientEncryptCloud

Use case: Multi-device sync, enterprise

Setup Guide

Step-by-step instructions for each deployment mode. Pick one and get started in minutes.

Fully managed on Neuromem Cloud. Memories are stored in hosted PostgreSQL + pgvector with automatic scaling, backups, and zero maintenance.

Go to neuromem.cloud, sign up, create an Agent, and generate an API Key.

Add to your MCP client

claude mcp add --scope user --transport http neuromem \
  https://api.neuromem.cloud/mcp/ \
  --header "Authorization: Bearer nm_sk_..."

ONE LLM Mode: Per-agent toggle in Dashboard > Agent settings. When on, your MCP client's LLM handles extraction — no server LLM needed. When off, the server's LLM handles extraction.