# MCP server

Add HARD to Claude Code or Claude Desktop as an MCP server, and the security checks it gives your assistant: messages, sites, calls and your Mac.

Source: https://hard.fyi/docs/mcp/

AI assistants that speak the Model Context Protocol, such as Claude, can use HARD’s checks. Add HARD to Claude Code:

```sh
claude mcp add hard -- /Applications/HARD.app/Contents/MacOS/hard-bridge mcp
```

Or to Claude Desktop, in its configuration file:

```json
{
  "mcpServers": {
    "hard": {
      "command": "/Applications/HARD.app/Contents/MacOS/hard-bridge",
      "args": ["mcp"]
    }
  }
}
```

## Tools

| Tool | What it does |
| --- | --- |
| `check_message` | Is this legit? for a text, email or pasted text. |
| `check_website` | One address: lookalike, reported phishing, domain age. |
| `check_call` | Is the person on the phone a scam, from yes-or-no answers. |
| `security_report` | HARD’s latest full check of this Mac. |
| `live_traffic` | Programs using the internet now, and who they talk to. Takes about 12 seconds. |
| `firewall_log` | Recent connections seen by the firewall, allowed or blocked. |
| `explain_program` | What a program is, who made it, and whether its signature is valid. |

What a tool returns goes to the assistant you’re using, under that assistant’s own terms.