# Lotor MCP server

> Creates a local, signed, and tamper-evident log of AI agent actions while providing a human-in-the-loop approval gate for consequential tasks.

- Repository: https://github.com/githubscum/lotor
- Page: https://hostedmcp.org/servers/githubscum/lotor
- Transport: stdio
- Stars: 3
- License: Apache-2.0
- Latest version: 1.0.2

## Install

### Claude Code

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

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

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

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

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

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

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

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

```toml
[mcp_servers.lotor]
command = "npx"
args = ["-y", "lotor-mcp"]
```

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

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

## FAQ

### What does Lotor actually do?

It writes a signed, tamper-evident log of agent actions, including files touched, messages sent, and costs incurred, and gates consequential actions behind a human signature.

### Is my data uploaded to a cloud service?

No, Lotor is local-first; the receipt is written to your machine and nowhere else, with no upload path or account.