Phishclean MCP README

The full upstream README, mirrored for reference. Install config, tool schemas, adoption signals and our own overview live on the Phishclean MCP listing page.

phishclean-mcp

npm license

MCP server for PhishClean security analysis tools. Works with Claude Desktop, Claude Code, Gemini CLI, and any MCP-compatible client.

Quick Start

Terminal
npx phishclean-mcp

Or add it to Claude Code in one line:

Terminal
claude mcp add phishclean -- npx -y phishclean-mcp

Remote server (Claude connector)

No install needed: PhishClean also runs as a remote MCP server at

Code
https://www.phishclean.com/mcp

In Claude (web, desktop or mobile): Customize → Connectors → + → Add custom connector, name it PhishClean, and paste that URL. No sign-in is needed. In Claude Code:

Terminal
claude mcp add --transport http phishclean https://www.phishclean.com/mcp

The remote server has the same tools as this package except check_password, so nobody is invited to type a real password into a chat. Every tool is read-only and runs the same scanner as the matching /api/v1 endpoint, in memory, without fetching anything. Limits: 1,000 tool calls an hour per IP without a key, or your plan's limit with Authorization: Bearer pk_live_….

Claude Desktop Setup

Add to your claude_desktop_config.json:

config.json
{
  "mcpServers": {
    "phishclean": {
      "command": "npx",
      "args": ["-y", "phishclean-mcp"]
    }
  }
}

Config file location:

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json

Tools

ToolDescription
check_linkCheck a URL for phishing signals (domain spoofing, suspicious TLDs, token leaks)
check_passwordAnalyze password strength (entropy, crack time, common patterns)
check_emailDetect phishing in email body text (urgency, threats, credential requests)
analyze_headersCheck email headers for SPF/DKIM/DMARC and spoofing
decode_jwtDecode JWT and flag security issues (expired, alg:none, sensitive data)
scan_secretsScan code for leaked API keys (AWS, Stripe, GitHub, Google, etc.)
scan_pageFull page scan — 12 phishing signals on HTML source

Examples

Once configured, ask Claude:

  • "Check if this URL is safe: https://paypal-secure.xyz/login"
  • "How strong is the password Summer2024!?"
  • "Is this email phishing? Dear customer, verify your account..."
  • "Scan this code for leaked secrets: const key = AKIA..."
  • "Decode this JWT: eyJhbGci..."

API Key (Optional)

Free tier: 30 requests/hour (no key needed).

For higher limits, get an API key at phishclean.com/developers:

TierRate LimitPrice
Free30 req/hr$0
Pro2,000 req/hr$19/mo
Business25,000 req/hr$49/mo
EnterpriseUnlimited$700/yr

Set your API key via environment variable:

config.json
{
  "mcpServers": {
    "phishclean": {
      "command": "npx",
      "args": ["-y", "phishclean-mcp"],
      "env": {
        "PHISHCLEAN_API_KEY": "pk_live_your_key_here"
      }
    }
  }
}

Or via shell:

bash
PHISHCLEAN_API_KEY=pk_live_your_key_here npx phishclean-mcp

API

All tools call the PhishClean API at https://www.phishclean.com/api/v1/.

Override the base URL with:

bash
PHISHCLEAN_API_URL=http://localhost:3000 npx phishclean-mcp

Privacy policy

Both the npm package and the remote server at https://www.phishclean.com/mcp handle data the same way:

  • Collected: only the arguments of each tool call (a URL, email text, headers, a token, source code or page HTML), plus your IP address or API key for rate limiting.
  • Use and storage: arguments are analysed in memory and discarded when the response is sent. They are not stored or logged. Rate-limit counters (IP or key hash, hourly count) are deleted after a few hours.
  • Third parties: nothing is shared or sold. The scanners make no outbound requests.
  • Retention: none for tool arguments; a few hours for rate-limit counters.
  • Contact: https://www.phishclean.com/contact. Full policy: https://www.phishclean.com/privacy

What this server sends, and where

This is a thin stdio client: each tool call is one HTTPS POST to https://www.phishclean.com/api/v1/<tool> carrying only the argument you passed (the URL, password, email text, headers, token, or source). Nothing is stored or logged server-side beyond rate limiting; passwords are analysed and discarded. The whole server is one file — read it before trusting it.

The PhishClean browser extension is a separate, fully local product whose source is also public: phishclean-extension.

Links

License

MIT — see LICENSE.