Llm Guard Privacy Gateway

Name: Llm Guard Privacy Gateway
Author: Amywith

@Amywith

リポジトリ

Llm Guard Privacy Gateway について

概要はまだありません

#llm-guard #privacy-gateway #security

基本情報

カテゴリ

AI とエージェント

トランスポート

stdio

公開者

Amywith

投稿者

zhongjiemei

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "llm-guard-mcp": {
      "command": "python",
      "args": [
        "-m",
        "spacy",
        "download",
        "zh_core_web_sm",
        "--index-url",
        "https://pypi.tuna.tsinghua.edu.cn/simple/"
      ]
    }
  }
}

ツール

ツールは検出されませんでした

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is Llm Guard Privacy Gateway?

A privacy protection gateway based on the MCP (Model Context Protocol) protocol. It detects and filters sensitive information in LLM applications—automatically masking Chinese names, phone numbers, emails, intercepting API keys and RSA private keys, and blocking custom blacklist words. It wraps these capabilities as MCP tools, returning JSON results.

How to use Llm Guard Privacy Gateway?

Install dependencies with pip install -r requirements.txt. Use the Python function sanitize_prompt_mcp or the LLMSanitizer class. To run as an MCP server, execute python mcp_server.py or use the provided Docker image.

Key features of Llm Guard Privacy Gateway?

PII masking: identifies Chinese names, phone numbers, and emails.
Key interception: detects API keys, Alibaba Cloud AccessKeys, and RSA private keys.
Commercial blocking: predefined blacklist words trigger an intercepted status.
MCP encapsulation: logic exposed as MCP tools returning structured JSON.

Use cases of Llm Guard Privacy Gateway?

Sanitizing user prompts before sending to an LLM to prevent data leakage.
Filtering sensitive information in enterprise chat applications.
Protecting against accidental exposure of API keys or private keys.
Enforcing content policy by blocking predefined blacklist terms.

FAQ from Llm Guard Privacy Gateway

What happens if spaCy model fails to download?

The system automatically falls back to built‑in regex detection. Accuracy may slightly decrease, but the server continues to run.

Does data processing require an internet connection?

No. Once dependencies are installed, all data filtering and detection is performed locally. The MCP server can run fully offline.

How can I customize the blacklist words?

Edit the LLMSanitizer class in llm_guard.py to modify the blacklist vocabulary or other detection rules.

What error handling mechanisms are in place?

All exceptions are caught gracefully. For example, if the presidio library fails, the system uses built‑in regex patterns without crashing the entire application.