MCPVet MCP server

Vet an MCP server, Claude Code skill, or plugin for unsafe behavior — shell execution, secret access, data exfiltration, prompt injection, remote code — before you install it, right inside your agent.

It's a thin, auditable client: all analysis runs server-side at mcpvet.com. This process holds no credentials, touches no local files, and makes exactly one outbound HTTPS call per scan.

Install (Claude Code)

claude mcp add mcpvet -- npx -y github:LorenzoLombardi111/factory-mcpvet

Or add it to your MCP config manually:

{
  "mcpServers": {
    "mcpvet": {
      "command": "npx",
      "args": ["-y", "github:LorenzoLombardi111/factory-mcpvet"]
    }
  }
}

Related MCP server: mcp-security-audit

Use

Ask your agent:

Before I install it, scan @some/mcp-server with MCPVet.

The scan_extension tool accepts:

• a GitHub repo URL — https://github.com/owner/repo

• an npm package name — @scope/pkg

• pasted skill / manifest text

It returns a graded verdict (clean → critical), the flagged findings with file and line, and a shareable report link.

Config

MIT licensed.