Skip to content

GlyphGuard

by glyphguard.mike-tusa.workers.dev · remote · no sign-in to list tools · v0.1.0 · published by dev.workers.mike-tusa.glyphguard

Find hidden Unicode (zero-width, bidi, tag payloads), homoglyphs and lookalike names in text.

Read only
CompareRegistry entry

3 tools: 3 only read.

Scanned Oct 22, 2026

No check matched anything in the tool definitions. No finding is not a safety guarantee. How the checks work.

claude mcp add --transport http glyphguard https://glyphguard.mike-tusa.workers.dev/mcp

Tools

Open a tool for its reasons and full description. Descriptions are the publisher's own words, shown as they are.

3 tools

Only reads 3Claude Code: allow
  • clean_textReturn the text with invisible characters, bidi controls, tag characters and control characters removed and unusual spaces turned into normal spaces.No findings

    Why it is in this group

    • The description's first verb is "return": a word for reading.
    Description, as the server wrote it
    Return the text with invisible characters, bidi controls, tag characters and control characters removed and unusual spaces turned into normal spaces. By default ZWJ/ZWNJ/VS16 are kept inside emoji sequences, between letters of joining scripts (Arabic, Persian) and after an Indic virama (keep_joiners=true); strict=true removes them everywhere. Optional: map lookalike letters to Latin (lookalikes) and apply NFKC normalisation (nfkc). Read-only. Same engine as GET /api/clean.

    Annotations the server declares

    • title: Clean hidden Unicode
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • compare_namesCheck whether a name, username, domain label or brand looks the same as any of up to 50 protected names (for example "pаypal" with a Cyrillic а vs "paypal"), using a UTS #39-style …No findings

    Why it is in this group

    • The name "compare_names" says "compare": a word for reading.
    Description, as the server wrote it
    Check whether a name, username, domain label or brand looks the same as any of up to 50 protected names (for example "pаypal" with a Cyrillic а vs "paypal"), using a UTS #39-style confusable skeleton (curated subset, case and accents ignored). Read-only. Same engine as GET /api/compare.

    Annotations the server declares

    • title: Compare names for lookalikes
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • scan_textFind invisible characters, bidi controls (Trojan Source), hidden Unicode tag payloads (decoded), variation-selector runs, control characters, unusual spaces, mixed-script words and…No findings

    Why it is in this group

    • The description's first verb is "find": a word for reading.
    Description, as the server wrote it
    Find invisible characters, bidi controls (Trojan Source), hidden Unicode tag payloads (decoded), variation-selector runs, control characters, unusual spaces, mixed-script words and lookalike letters. ZWJ/ZWNJ/VS16 in legitimate context (emoji sequences, Persian/Arabic joining, Indic virama) are reported as info joiner_in_context and do not raise risk; elsewhere they stay medium. Lone surrogates are flagged. Returns a risk level (none/low/medium/high), counts, and findings with UTF-16 offsets and code points. Read-only. Same engine as GET /api/scan.

    Annotations the server declares

    • title: Scan text for hidden Unicode
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false

This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.

Install

Pinned to version 0.1.0. Values in angle brackets are yours to fill in.

claude mcp add --transport http glyphguard https://glyphguard.mike-tusa.workers.dev/mcp

Run in a terminal.

Claude Code permissions

Allow the reads, ask before a change, deny the deletes. Paste into .claude/settings.json.

.claude/settings.json
{
  "permissions": {
    "allow": [
      "mcp__glyphguard__scan_text",
      "mcp__glyphguard__clean_text",
      "mcp__glyphguard__compare_names"
    ],
    "ask": [],
    "deny": []
  }
}

Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. It assumes the server is installed as glyphguard. Review it before you use it.