# SuperMemory MCP server

> Enables AI agents to capture distilled lessons from failures and corrections to improve performance through a closed-loop learning cycle.

- Repository: https://github.com/YashvantHange/SuperMemory
- Page: https://hostedmcp.org/servers/yashvanthange/supermemory
- Transport: stdio
- Stars: 2
- License: MIT
- Latest version: 0.2.5

## Install

### Claude Code

```sh
claude mcp add supermemory -- uvx supermemory-agent
```

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

```json
{
  "mcpServers": {
    "supermemory": {
      "command": "uvx",
      "args": [
        "supermemory-agent"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "supermemory": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "supermemory-agent"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"supermemory","type":"stdio","command":"uvx","args":["supermemory-agent"]}'
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "supermemory": {
      "command": "uvx",
      "args": [
        "supermemory-agent"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add supermemory -- uvx supermemory-agent
```

### Codex config.toml (`~/.codex/config.toml`)

```toml
[mcp_servers.supermemory]
command = "uvx"
args = ["supermemory-agent"]
```

### Gemini CLI

```sh
gemini mcp add -s user supermemory uvx -- supermemory-agent
```

### Gemini settings.json (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "supermemory": {
      "command": "uvx",
      "args": [
        "supermemory-agent"
      ]
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `SUPERMEMORY_STORAGE_PATH` | no | no | Local directory for .supermemory file storage |
| `UALL_DATA_DIR` | no | no | Legacy alias for SUPERMEMORY_STORAGE_PATH |

## FAQ

### What does SuperMemory capture?

It captures distilled lessons from failures and corrections rather than full conversation transcripts.
