MCP guide · Security Proxy

@guardbee/mcp-security-proxy is a transparent security layer in front of any MCP server. It catches prompt-injection patterns, masks PII in responses, and audits every hop — no code changes in the target server.
What does it do?
Use it before exposing filesystem, DB, or third-party MCP tools to Claude/Cursor. Requests pass the proxy first; hostile instructions are blocked or warned.
Who is it for?
- Teams connecting multiple MCP servers to Claude or Cursor
- Security owners reducing prompt-injection risk
- Product teams preventing PII leakage in tool responses
What it is not
- Acting as a business-logic MCP by itself (it wraps another server)
- Replacing a network WAF or API gateway
Features
| Feature | Description |
|---|---|
| Prompt injection defense | 16+ patterns including jailbreaks and hidden commands |
| PII masking | IDs, IBANs, emails, JWTs, API keys in responses |
| Block / warn | Per-interceptor mode |
| Audit log | Console or file |
| Zero code change | Launch the target after — |
MCP tools
| Tool | Description |
|---|---|
proxy_status | Active interceptors and blocked-request stats |
Target tools | Downstream MCP tools pass through transparently |
Install with npm
Global install (handy for Claude Desktop / Cursor):
npm install -g @guardbee/mcp-security-proxyor via npx (recommended with -y in config):
npx -y @guardbee/mcp-security-proxyThe proxy runs the command after -- as the target MCP — e.g. wrap the filesystem server.
Claude Desktop configuration
Config file paths:
- macOS
~/Library/Application Support/Claude/claude_desktop_config.json - Windows
%APPDATA%\Claude\claude_desktop_config.json - Linux
~/.config/Claude/claude_desktop_config.json
Merge this block, then restart Claude Desktop:
{
"mcpServers": {
"secure-filesystem": {
"command": "npx",
"args": [
"-y",
"@guardbee/mcp-security-proxy",
"--",
"npx",
"-y",
"@modelcontextprotocol/server-filesystem",
"/tmp"
]
}
}
}Cursor and other MCP clients
In Cursor, add the same command / args / env under Settings → MCP. stdio MCP servers use the same contract as Claude Desktop.
Example prompts in Claude
- “Show recent blocked requests via proxy_status”
- “List /tmp through this secured MCP”
- “Confirm emails are masked in responses”
Related links
See surface risk with Guardbee
MCP servers speed up security work in chat. Start a 14-day free trial for brand scans and BeeAI.
Frequently asked questions
Which MCP servers work?
Most stdio-launched MCP servers; you supply the command after –.
Block or warn?
Configure per interceptor — prefer block in sensitive environments.
Performance?
Pattern checks are cheap; watch audit file I/O at high volume.