Quick Start

1. Install

To use local embeddings, pull an Ollama model first (~270MB, free, no GPU required):

ollama pull nomic-embed-text

Then install memtomem:

uv tool install memtomem          # or: pipx install memtomem

To also use the STM proxy:

uv tool install memtomem-stm      # or: pip install memtomem-stm

No GPU? Pick OpenAI in the setup wizard — see the Embeddings guide.

2. Setup

Run the interactive 8-step wizard:

mm init                            # interactive setup
mm init -y                         # auto-accept defaults (for CI)

3. Connect your MCP client

Claude Code

claude mcp add memtomem -s user -- memtomem-server

With STM proxy:

claude mcp add memtomem-stm -s user -- memtomem-stm

Cursor / Windsurf / Claude Desktop

Register memtomem-server or memtomem-stm as a stdio MCP server in your editor’s MCP settings. See MCP Client Setup for details.

4. Use

Use memory through your AI agent:

Tell your agent	MCP tool called
”Check server status”	mem_status
”Index my ~/notes folder”	mem_index(path="~/notes")
”Search for deployment”	mem_search(query="deployment checklist")
”Remember this insight”	mem_add(content="...", tags=["ops"])

Packages

Package	Description
memtomem	Core — MCP server, CLI (mm), Web UI, hybrid search, storage
memtomem-stm	STM proxy — proactive memory surfacing via tool interception

The two packages have no Python-level dependency on each other. All communication between STM and LTM happens over the MCP protocol — they can be deployed and upgraded independently.