# Minnimemory

> Minnimemory is a free MCP server for knowledge and memory. Splits Claude Code and Codex memory files into always-loaded and on-demand parts. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Free
- **Platforms:** CLI, macOS, Windows, Linux
- **Website:** https://minniai.com/minnimemorymcp
- **Repository:** https://github.com/minniai-com/minnimemory-plugin
- **GitHub stars:** 0
- **MCP Registry name:** io.github.MinniAI-com/minnimemory
- **Categories:** knowledge and memory, developer tools
- **Listing updated:** 2026-11-14

## About

Memory files such as CLAUDE.md, AGENTS.md and a memory folder are re-read on every turn, and most of their content has nothing to do with the task at hand. This server audits those files and compiles them into a small always-on section plus separate files that load only when a prompt needs them.

The workflow is four commands. `init` maps what gets loaded and previews the split without changing anything; `apply` makes the change after backing up the originals; `undo` restores them byte for byte; `doctor` re-checks later for drift. In Claude Code, apply and undo return a command that runs through your approval prompt. The developer reports one sample map going from 12,511 to 6,699 tokens per turn and, in its best Claude Code run, a session cost of $0.83 falling to $0.57; the median of 13 runs was 15.9% cheaper. These are the vendor's own figures from its own memory folder.

It runs entirely on your computer with no model inside it, no API key and no network traffic. It supports Claude Code and Codex only. The plugin also adds hooks that inject the one matching memory section before each prompt and warn when memory drifted. The pricing is free during early access and the rule set is proprietary.

## Features

- **Memory map:** Lists which files load every turn and which content could move to on-demand.
- **Lossless apply:** Backs up originals first and keeps every line.
- **Undo:** Restores the original files byte for byte.
- **Drift check:** doctor shows what changed since the last apply and a re-apply plan.
- **Prompt hook:** Adds only the memory section matching the prompt, marked as data.
- **Local only:** Runs with no model, no API key and no network traffic.

## Good for

- Cut the token cost of a bloated CLAUDE.md
- Trim an oversized AGENTS.md for Codex
- Audit which memory files an agent loads on each turn
- Keep memory tidy as notes grow week by week

## Install

### Claude Code (Terminal)

```sh
claude mcp add minnimemory -- npx -y minnimemory
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "minnimemory": {
      "command": "npx",
      "args": [
        "-y",
        "minnimemory"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "minnimemory": {
      "command": "npx",
      "args": [
        "-y",
        "minnimemory"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "minnimemory": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "minnimemory"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "minnimemory": {
      "command": "npx",
      "args": [
        "-y",
        "minnimemory"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "minnimemory": {
      "command": "npx",
      "args": [
        "-y",
        "minnimemory"
      ]
    }
  }
}
```

## Tools

- `init`: Maps the memory files your agent loads and previews the split.
- `doctor`: Re-checks memory for drift and shows a re-apply plan.
- `apply`: Splits memory into always-on and on-demand parts, returning a command for approval.
- `undo`: Restores the original files.

## FAQ

### Does it delete anything?

No. The site says every line is kept and the originals are backed up before apply.

### Which agents work with it?

Claude Code and Codex. Other tools are not supported yet.

### What does it cost?

Free during early access, per the website. The rule set is proprietary.

### Does it send data off my machine?

No. The tools send nothing over the internet.

### What does it require?

Node.js 18.3 or later.

### Is Minnimemory free?

Yes. Minnimemory is free to use.

### How do I add Minnimemory to Claude Code?

Run this command in your terminal: claude mcp add minnimemory -- npx -y minnimemory

### What are the best alternatives to Minnimemory?

The closest alternatives to Minnimemory are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Nautilus Compass Memory](https://ai-powered.com/mcp/nautilus-compass-memory): Nautilus Compass Memory is an open-source MCP server for knowledge and memory.

---
Source: https://ai-powered.com/mcp/minnimemory
