Skip to main content
Glama
hebrew-scam-shield-mcp

hebrew-scam-shield-mcp

by benziRaz

Hebrew Scam Shield ๐Ÿ›ก๏ธ

Explainable scam and phishing detection for Hebrew messages (SMS, WhatsApp, e-mail). Zero dependencies, a command-line tool, a Python library, and an MCP server so AI assistants can check messages for you.

ืขื‘ืจื™ืช โ†“

Fake "your parcel is held" texts, fake government and bank alerts, fake emergency-app links: these messages reach everyone, but they hurt most the people who are least likely to have help at hand: elderly people, new internet users, and communities that are under-served by mainstream security tools. Most scam filters are built for English. Hebrew Scam Shield is built for Hebrew, and every verdict comes with a plain-language reason, so the person (or the assistant) can understand why a message looks dangerous.

Quick start

pip install "hebrew-scam-shield[mcp] @ git+https://github.com/benziRaz/hebrew-scam-shield"

hebrew-scam-shield "ื“ื•ืืจ ื™ืฉืจืืœ: ื”ื—ื‘ื™ืœื” ืฉืœืš ืœื ื ื™ืชื ืช ืœืžืกื™ืจื”. ืืฉืจ ื›ืชื•ื‘ืช ืชื•ืš 12 ืฉืขื•ืช http://israelpost-il.top/x"
HIGH RISK  (score 90/100)
  +25  Claims a parcel/delivery problem (address, payment, customs) - a very common lure.
  +15  Creates artificial urgency or threatens a deadline.
  +15  Link uses a top-level domain that is rarely used by genuine services.  [http://israelpost-il.top/x]
  +5  Link is not encrypted (http).  [http://israelpost-il.top/x]
  +30  Domain contains a known brand/institution name but is not that organisation's domain.  [http://israelpost-il.top/x]

Use --lang he for Hebrew explanations and --json for machine-readable output. Exit codes: 0 low, 1 suspicious, 2 high.

As a library

from hebrew_scam_shield import analyze

result = analyze(message_text)
result.level      # "low" | "suspicious" | "high"
result.score      # 0-100
result.reasons    # list of Reason(id, weight, en, he, detail)

As an MCP server (for Claude and other assistants)

{
  "mcpServers": {
    "hebrew-scam-shield": { "command": "hebrew-scam-shield-mcp" }
  }
}

The server exposes one tool, check_message(text). Ask your assistant: "Is this message a scam?" and paste the text.

Related MCP server: ScamCheck MCP Server

What it looks for

Signal

Example

Parcel / customs / address problem

"ื”ื—ื‘ื™ืœื” ืœื ื ื™ืชื ืช ืœืžืกื™ืจื”, ืืฉืจ ื›ืชื•ื‘ืช"

Artificial urgency

"ืชื•ืš 12 ืฉืขื•ืช", "ืื—ืจืช ื”ื—ืฉื‘ื•ืŸ ื™ื™ื—ืกื"

Government / bank impersonation with an account problem

"ืคืขื™ืœื•ืช ื—ืจื™ื’ื” ื‘ื—ืฉื‘ื•ืŸ" + link

Requests for verification codes, passwords, card data

"ื”ื–ืŸ ืืช ืงื•ื“ ื”ืื™ืžื•ืช"

Fake emergency or "security update" app installs

"ื”ื•ืจื“ ืืช ื”ืืคืœื™ืงืฆื™ื”"

Prize and voucher lures

"ื–ื›ื™ืช ื‘ืฉื•ื‘ืจ"

Wallet-linking tricks

"ืœืงืฉืจ Apple Pay ืœืฉื—ืจื•ืจ ื”ื—ื‘ื™ืœื”"

"Reply Y to activate the link"

"ื”ืฉื‘ Y"

Suspicious links

shorteners, look-alike brand domains, odd TLDs, raw IPs, @ tricks, .apk files, punycode

Evasion

zero-width characters, niqqud

Rules live in rules.py as readable data (weight + English/Hebrew explanation + regex groups). Sources are listed in docs/PATTERNS.md.

Honest limitations

  • It is a heuristic, not a guarantee. A low score does not prove a message is safe; a high score does not prove it is malicious.

  • Rules are public, so determined scammers can study them. Use this as one layer of defence, not the only one.

  • The list of official domains is short and non-exhaustive. Pull requests that add verified domains are very welcome.

  • It was built from public warnings, not from a private corpus of real messages, so expect false positives and gaps.

  • Not affiliated with the Israel National Cyber Directorate. If you receive a suspicious message in Israel, you can report it to the national 119 hotline.

Contributing

New scam patterns, official domains and false-positive reports are the most valuable contributions. See CONTRIBUTING.md. Never paste real personal data into issues: redact names, phone numbers and ID numbers first.

pip install -e ".[mcp,dev]"
pytest

License: MIT


ืขื‘ืจื™ืช

ื–ื™ื”ื•ื™ ื”ื•ื ืื•ืช ื•ื”ื•ื“ืขื•ืช ืคื™ืฉื™ื ื’ ื‘ืขื‘ืจื™ืช, ืขื ื”ืกื‘ืจ ืœื›ืœ ื”ื—ืœื˜ื”. ื”ื›ืœื™ ืขื•ื‘ื“ ืขืœ SMS, ื•ื•ืื˜ืกืืค ื•ืžื™ื™ืœ, ื‘ืœื™ ืชืœื•ื™ื•ืช ื—ื™ืฆื•ื ื™ื•ืช, ื•ืžื’ื™ืข ื›ืฉื•ืจืช ืคืงื•ื“ื”, ื›ืกืคืจื™ื™ื” ื‘ืคื™ื™ืชื•ืŸ ื•ื›ืฉืจืช MCP ืฉืžืืคืฉืจ ืœืขื•ื–ืจื™ AI ืœื‘ื“ื•ืง ื”ื•ื“ืขื•ืช ืขื‘ื•ืจืš.

ื”ื•ื“ืขื•ืช ืขืœ "ื—ื‘ื™ืœื” ืฉื ืชืงืขื”", ื”ืชืจืื•ืช ืžื–ื•ื™ืคื•ืช ืžื”ื‘ื ืง ืื• ืžื”ืžื“ื™ื ื” ื•ืงื™ืฉื•ืจื™ื ืœืืคืœื™ืงืฆื™ื•ืช ื—ื™ืจื•ื ืžื–ื•ื™ืคื•ืช ืžื’ื™ืขื•ืช ืœื›ื•ืœื, ืื‘ืœ ื”ื ื–ืง ื’ื“ื•ืœ ื‘ืžื™ื•ื—ื“ ืืฆืœ ืžื™ ืฉืื™ืŸ ืœื™ื“ื• ืžื™ ืฉื™ืขื–ื•ืจ: ืงืฉื™ืฉื™ื, ื’ื•ืœืฉื™ื ื—ื“ืฉื™ื ื•ืงื”ื™ืœื•ืช ืฉื›ืœื™ ื”ืื‘ื˜ื—ื” ื”ืจื’ื™ืœื™ื ืœื ื‘ื ื•ื™ื™ื ื‘ืฉื‘ื™ืœืŸ. ืจื•ื‘ ื”ืžืกื ื ื™ื ื ื›ืชื‘ื• ืœืื ื’ืœื™ืช. ื”ื›ืœื™ ื”ื–ื” ื ื‘ื ื” ืœืขื‘ืจื™ืช, ื•ื›ืœ ืชื•ืฆืื” ืžืœื•ื•ื” ื‘ื”ืกื‘ืจ ื‘ืจื•ืจ ืœืžื” ื”ื”ื•ื“ืขื” ื ืจืื™ืช ืžืกื•ื›ื ืช.

ื”ืชืงื ื” ื•ืฉื™ืžื•ืฉ

pip install "hebrew-scam-shield[mcp] @ git+https://github.com/benziRaz/hebrew-scam-shield"
hebrew-scam-shield --lang he "ื“ื•ืืจ ื™ืฉืจืืœ: ื”ื—ื‘ื™ืœื” ืฉืœืš ืœื ื ื™ืชื ืช ืœืžืกื™ืจื”. ืืฉืจ ื›ืชื•ื‘ืช ืชื•ืš 12 ืฉืขื•ืช http://israelpost-il.top/x"

ืงื•ื“ ื™ืฆื™ืื”: 0 ืกื™ื›ื•ืŸ ื ืžื•ืš, 1 ื—ืฉื•ื“, 2 ืกื™ื›ื•ืŸ ื’ื‘ื•ื”.

ืžื’ื‘ืœื•ืช

  • ื–ื• ื”ืขืจื›ื” ืœืคื™ ื›ืœืœื™ื ื•ืœื ื•ื“ืื•ืช. ืฆื™ื•ืŸ ื ืžื•ืš ืœื ืžื•ื›ื™ื— ืฉื”ื”ื•ื“ืขื” ื‘ื˜ื•ื—ื”, ื•ืฆื™ื•ืŸ ื’ื‘ื•ื” ืœื ืžื•ื›ื™ื— ืฉื”ื™ื ื–ื“ื•ื ื™ืช.

  • ื”ื›ืœืœื™ื ืคื•ืžื‘ื™ื™ื, ื•ืœื›ืŸ ื ื•ื›ืœื™ื ื ื—ื•ืฉื™ื ื™ื›ื•ืœื™ื ืœืœืžื•ื“ ืื•ืชื. ื›ื“ืื™ ืœื”ืฉืชืžืฉ ื‘ื›ืœื™ ื›ืื—ืช ืžืฉื›ื‘ื•ืช ื”ื”ื’ื ื” ื•ืœื ื›ืฉื›ื‘ื” ื™ื—ื™ื“ื”.

  • ืจืฉื™ืžืช ื”ื“ื•ืžื™ื™ื ื™ื ื”ืจืฉืžื™ื™ื ืงืฆืจื” ื•ื—ืœืงื™ืช. ืชื•ืกืคื•ืช ืžืื•ืžืชื•ืช ื™ืชืงื‘ืœื• ื‘ื‘ืจื›ื”.

  • ื”ื›ืœื™ ืœื ืงืฉื•ืจ ืœืžืขืจืš ื”ืกื™ื™ื‘ืจ ื”ืœืื•ืžื™. ืžื™ ืฉืงื™ื‘ืœ ื”ื•ื“ืขื” ื—ืฉื•ื“ื” ื‘ื™ืฉืจืืœ ื™ื›ื•ืœ ืœื“ื•ื•ื— ืœืงื• 119.

ืชืจื•ืžื” ืœืคืจื•ื™ืงื˜

ื“ืคื•ืกื™ ื”ื•ื ืื” ื—ื“ืฉื™ื, ื“ื•ืžื™ื™ื ื™ื ืจืฉืžื™ื™ื ืžืื•ืžืชื™ื ื•ื“ื™ื•ื•ื—ื™ื ืขืœ ื”ืชืจืื•ืช ืฉื•ื•ื ื”ื ื”ืชืจื•ืžื” ื”ื›ื™ ื—ืฉื•ื‘ื”. ืคืจื˜ื™ื ื‘-CONTRIBUTING.md. ืืœ ืชื“ื‘ื™ืงื• ืžื™ื“ืข ืื™ืฉื™ ืืžื™ืชื™ ื‘-issues: ืžื—ืงื• ืฉืžื•ืช, ื˜ืœืคื•ื ื™ื ื•ืžืกืคืจื™ ื–ื”ื•ืช.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelogโ€ฆ

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      D
      maintenance
      AI-powered threat verification across 6 channels: phone numbers, URLs, text messages, emails, documents, and QR codes. Analyzes threats against 10M+ intelligence records from FTC, FCC, URLhaus, ThreatFox, and community reports, returning risk scores, verdicts, and detailed signals.
      10
      1
      MIT
    • A
      license
      A
      quality
      C
      maintenance
      Scans suspicious messages, URLs, and text for scams inside any MCP-compatible AI assistant. No signup or API key needed for anonymous use.
      1
      60 npm
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables AI assistants to autonomously detect scams by checking URLs, messages, ABNs, and phone numbers using Australian-specific threat intelligence and privacy-preserving techniques.
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Enables AI assistants to check whether a crypto token, link, or forwarded "gem" message is a scam by running contract rules, live test trades, creator history, and market data across 64 networks, returning plain-English or Pidgin verdicts. Also supports scanning messages, reviewing a token's past checks, and explaining individual findings.
      5
      MIT