# model-scanner MCP server

> Scans machine learning model files like PyTorch, safetensors, and GGUF for supply-chain risks and malicious code execution vulnerabilities.

- Repository: https://github.com/GuardBee/guardbee-mcp
- Page: https://hostedmcp.org/servers/guardbee/model-scanner
- Transport: stdio
- Stars: 2
- License: MIT
- Latest version: 0.1.2

## Install

### Claude Code

```sh
claude mcp add model-scanner -- npx -y @guardbee/mcp-model-scanner
```

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

```json
{
  "mcpServers": {
    "model-scanner": {
      "command": "npx",
      "args": [
        "-y",
        "@guardbee/mcp-model-scanner"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "model-scanner": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@guardbee/mcp-model-scanner"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"model-scanner","type":"stdio","command":"npx","args":["-y","@guardbee/mcp-model-scanner"]}'
```

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

```json
{
  "mcpServers": {
    "model-scanner": {
      "command": "npx",
      "args": [
        "-y",
        "@guardbee/mcp-model-scanner"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add model-scanner -- npx -y @guardbee/mcp-model-scanner
```

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

```toml
[mcp_servers.model-scanner]
command = "npx"
args = ["-y", "@guardbee/mcp-model-scanner"]
```

### Gemini CLI

```sh
gemini mcp add -s user model-scanner npx -- -y @guardbee/mcp-model-scanner
```

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

```json
{
  "mcpServers": {
    "model-scanner": {
      "command": "npx",
      "args": [
        "-y",
        "@guardbee/mcp-model-scanner"
      ]
    }
  }
}
```

## Tools (4)

- `scan_file`
- `scan_directory`
- `scan_hf_repo`
- `list_patterns`

## FAQ

### Can I scan a Hugging Face repository directly?

Yes, you can use the scan_hf_repo tool to scan every model file in a Hugging Face repo at a specific pinned commit.

### Does the scanner load the entire model into memory?

No, it uses bounded reads and specific extraction methods, such as reading the ZIP central directory for PyTorch files, to avoid loading multi-gigabyte checkpoints fully into memory.

### Is telemetry enabled by default?

No, telemetry is turned off by default.
