
MCP Guard 保护您的 MCP 客户端免受即时注入攻击等多种攻击。
概览
一个MCP(模型上下文协议)护栏,内置AI驱动的管理功能,将多个MCP服务器聚合到一个安全接口中。 MCP Guardrail 为您的 MCP(模型上下文协议)设置提供 AI 驱动的安全性和简便配置。它自动检测您现有的 MCP 配置文件,并添加具备智能审核功能的保护层。 - 防止提示注入攻击 - CLI 工具检测并更新 Cursor、Claude Desktop 和 Claude Code 的 MCP 配置文件 - 支持 本地和远程 MCP 服务器 - 工具、提示和资源自动前缀并开放 最简单的入门方式是使用通用分析CLI工具: 这会自动更新你的MCP配置文件,并设置护栏。 无需安装!
概述(AI 翻译)
一个MCP(模型上下文协议)护栏,内置AI驱动的管理功能,将多个MCP服务器聚合到一个安全接口中。
概述
MCP Guardrail 为您的 MCP(模型上下文协议)设置提供 AI 驱动的安全性和简便配置。它自动检测您现有的 MCP 配置文件,并添加具备智能审核功能的保护层。
主要特点:
- AI驱动的管理以防止即时注入攻击
- 自动配置 - CLI 工具检测并更新 MCP 配置文件,适用于光标、Claude Desktop 和 Claude 代码
- 双重连接 - 支持本地和远程MCP服务器
- 透明代理 - 工具、提示和资源自动前缀并开放
快速启动
最简单的入门方式是使用通用分析CLI工具:
这会自动更新你的MCP配置文件,并设置护栏。
工作原理
用途
与NPX的直接使用
无需安装!直接在你的光标或Claude桌面MCP配置中使用:
独立用途
配置
本地服务器(Stdio)
本页由 AI 翻译,技术细节请以原文为准。
README
以下为原文 README 节选,完整内容请查看原文。
MCP Guardrail
An MCP (Model Context Protocol) guardrail with built-in AI-powered moderation that aggregates multiple MCP servers into one secure interface.
Overview
The MCP Guardrail provides AI-powered security and easy configuration for your MCP (Model Context Protocol) setup. It automatically detects your existing MCP configuration files and adds a protective layer with intelligent moderation capabilities.
Key features:
- AI-powered moderation to prevent prompt injection attacks
- Automatic configuration - CLI tool detects and updates MCP config files for Cursor, Claude Desktop, and Claude Code
- Dual connectivity - supports both local and remote MCP servers
- Transparent proxying - tools, prompts, and resources are automatically prefixed and made available
Quick Start
The easiest way to get started is using the General Analysis CLI tool:
# Install the CLI tool
pip3 install generalanalysis==0.1.7
# Login to your account
ga login
# Configure MCP settings for Cursor, Claude Desktop, and Claude Code
ga configure
This will automatically update your MCP configuration files with the guardrail setup.
How It Works
graph LR
A[MCP Client] --> B[MCP Guard]
B --> C[MCP Server]
C --> D[Tool Output]
D --> B
B --> E{GA Guardrail}
E -->|Safe| A
E -->|Blocked| F[Alert]
Usage
Direct Usage with npx
No installation required! Use directly in your Cursor or Claude Desktop MCP configuration:
{
"mcpServers": {
"protected_server": {
"command": "npx",
"args": [
"-y",
"@general-analysis/mcp-guard",
"[{\"name\":\"server1\",\"command\":\"path/to/server\",\"args\":[\"arg1\"]}]"
]
}
}
}
Standalone Usage
npx -y @general-analysis/mcp-guard '[{"name":"server1","command":"path/to/server","args":["arg1"]}]'
Configuration
Local Servers (Stdio)
For local MCP servers that communicate via stdio:
{
"mcpServers": {
"protected_server": {
"command": "npx",
"args": [
"-y",
"@general-analysis/mcp-guard",
"[{\"name\":\"my-local-server\",\"command\":\"node\",\"args\":[\"path/to/server.js\"]}]"
],
"env": {
"API_KEY": "your-general-analysis-api-key",
"ENABLE_GUARD_API": "true"
}
}
}
}
Remote Servers (HTTP/SSE)
For remote MCP servers accessible via HTTP or Server-Sent Events:
{
"mcpServers": {
"protected_server": {
"command": "npx",
"args": [
"-y",
"@general-analysis/mcp-guard",
"[{\"name\":\"my-remote-server\",\"url\":\"https://api.example.com/mcp\"}]"
],
"env": {
"API_KEY": "your-general-analysis-api-key",
"ENABLE_GUARD_API": "true"
}
}
}
}
Environment Variables
API_KEY- Your General Analysis API key for the moderation service
安装
npx -y @general-analysis/mcp-guard [{\"name\":\"server1\ \"command\":\"path/to/server\ \"args\":[\"arg1\配置
{
"mcpServers": {
"protected_server": {
"command": "npx",
"args": [
"-y",
"@general-analysis/mcp-guard",
"[{\"name\":\"server1\",\"command\":\"path/to/server\",\"args\":[\"arg1\"]}]"
]
}
}
}