# coding-agent-audit MCP server

> Scans machines and MCP configurations for supply-chain backdoors, risky permissions, and malicious tool descriptions in AI coding assistants.

- Repository: https://github.com/stackscribe-ai/coding-agent-audit
- Page: https://hostedmcp.org/servers/stackscribe-ai/coding-agent-audit
- Transport: stdio
- Stars: 0
- Latest version: 0.11.0

## Install

### Claude Code

```sh
claude mcp add coding-agent-audit -- uvx coding-agent-audit
```

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

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

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

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

### VS Code CLI

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

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

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

### Codex

```sh
codex mcp add coding-agent-audit -- uvx coding-agent-audit
```

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

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

### Gemini CLI

```sh
gemini mcp add -s user coding-agent-audit uvx -- coding-agent-audit
```

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

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

## FAQ

### What does this tool scan for?

It scans for known indicators of compromise (IOCs) from supply-chain attacks and risky configuration patterns like unpinned MCP packages or overly broad permissions.

### Can I run this as an MCP server?

Yes, you can add it to your client's MCP config to expose tools like scan_for_known_backdoors and scan_mcp_config directly to your AI assistant.

### Does it check for hallucinated package names?

Yes, the --check-deps flag performs a phantom-dependency scan to see if package names in your manifest exist on npm or PyPI to prevent slopsquatting.
