Skip to main content
Glama
miketusa-ctrl

SmuggleCheck

by miketusa-ctrl
OverviewSchemaSourceRelated ServersDiscussionsAdmin
JavaScript
Remote

SmuggleCheck

Find hidden prompt-injection text: ASCII smuggling, zero-width/emoji stego, bidi, homoglyphs. One layer, not a full defense: it does not judge prompt injection written in plain, visible words.

SmuggleCheck is not affiliated with, endorsed or sponsored by Meta, OpenAI, Anthropic, Google, Mistral AI, DeepSeek, Alibaba (Qwen), the Unicode Consortium or any other AI vendor, model vendor or MCP client maker.

Live service

Related MCP server: PromptWall MCP Server

Use it from an AI assistant (MCP)

SmuggleCheck is a remote MCP server (Streamable HTTP) at https://smugglecheck.mike-tusa.workers.dev/mcp. Tools: scan_text, compare_strings.

Cursor, Cline and Windsurf accept this config:

{
  "mcpServers": {
    "smugglecheck": {
      "type": "http",
      "url": "https://smugglecheck.mike-tusa.workers.dev/mcp"
    }
  }
}

Claude Desktop: open Settings > Connectors, choose "Add custom connector" and paste https://smugglecheck.mike-tusa.workers.dev/mcp. Or, for clients that only support local (stdio) servers, use the mcp-remote bridge: npx mcp-remote https://smugglecheck.mike-tusa.workers.dev/mcp.

VS Code uses a servers key instead of mcpServers.

The server.json file is the MCP Registry manifest for this server.

Run your own copy

Requires Node.js 22+ and a Cloudflare account.

npm install
npx wrangler@4 dev      # local dev server

wrangler.toml is a template: create your own KV namespaces / D1 databases / rate-limit namespaces and replace the <PLACEHOLDER> values before npx wrangler@4 deploy. Any secrets the Worker reads go in wrangler secret put (or .dev.vars locally), never in the repo.

Optional Worker secrets (leave unset if you do not need them): GLAMA_CLAIM (your glama.ai connector claim token, served at /.well-known/glama.json); VISITOR_SALT (a random salt for the privacy-friendly daily unique-visitor counter; while it is unset, no visitors are counted); MCP_REGISTRY_AUTH (your MCP Registry HTTP-domain proof, v=MCPv1; k=ed25519; p=<public key>, served at /.well-known/mcp-registry-auth; 404 while unset).

Third-party data

The code in this repository is licensed under MIT. The homoglyph data file src/lib/confusables-data.js is NOT MIT: it is generated from Unicode confusables.txt (UTS #39 18.0.0), © Unicode, Inc., and is licensed under the Unicode License v3 (https://www.unicode.org/license.txt). See NOTICE.

License

Code: MIT (c) Digital Promo Hub, see LICENSE. Homoglyph data file src/lib/confusables-data.js: Unicode License v3, © Unicode, Inc., see NOTICE.

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityMaintained
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

  • PromptWall MCP Server

    vtino17
    A
    license
    Not graded
    quality
    A
    maintenance
    Enables scanning LLM prompts and responses for prompt injection, jailbreaks, PII leakage, secret leakage, and other malicious content using deterministic rules, returning verdicts and safe redacted text.
    1
    MIT

  • genpark-agent-prompt-injection

    alphaparkinc
    A
    license
    Not graded
    quality
    B
    maintenance
    Enables MCP clients and autonomous agents to screen incoming prompts for system overrides, delimiter breakouts, and encoded payload evasion, helping block prompt injection and jailbreak attempts before execution.
    7
    MIT

  • MCP Injection Guard

    KaryawanSurga
    A
    license
    Not graded
    quality
    A
    maintenance
    Enables users to scan untrusted text, files, and directory trees for prompt-injection patterns and receive severity-ranked, explainable verdicts before content reaches an agent. It runs locally and returns clean, review, or blocked verdicts for safer agent input.
    1
    MIT