Shield MCP Server

Detect tool poisoning, exfiltration channels, and cross-origin escalations in MCP configs.

Data last scanned today · Reviewed today

23★ 554TypeScriptMIT

Overview

Shield MCP Server acts as an auditing tool for your Model Context Protocol environment. By scanning your configuration files, it identifies security vulnerabilities such as prompt injection risks, unauthorized sensitive file access, and potential data exfiltration channels.

Our verdict

This tool carries a high risk level with a security score of 23/100, primarily due to it being a community-maintained project that has not undergone editorial review. With no updates in over 450 days and reliance on a single contributor, users should exercise extreme caution. It is suitable for those experimenting with security analysis of their local MCP setups, but should not be treated as a definitive security solution.

Identifies prompt injection and sensitive file access vulnerabilities
Supports custom config paths and client identity spoofing
Includes a safe-list feature to exclude specific servers
Contains only 12 direct dependencies

Project has not received a commit in 458 days
Lacks official vendor maintenance and editorial review
High bus-factor risk with only one contributor

Setup

Claude Desktop

{
  "mcpServers": {
    "mcp-shield": {
      "args": [
        "-y",
        "mcp-shield"
      ],
      "command": "npx"
    }
  }
}

Paste into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), then restart Claude Desktop. Some servers require API keys or tokens as environment variables — check the project's README and add an "env" object if needed.

No read-only mode — running this grants full read/write access.
No built-in authentication — anyone with access to this config can invoke every tool it exposes.

Compatibility

Client	Local	Docker	Remote	Read-only
Claude Desktop
Cursor
VS Code
Windsurf
ChatGPT

Frequently asked questions

›What locations does MCP-Shield scan by default?

If no path is provided, the tool automatically scans standard locations including ~/.config/.mcp, ~/Library/Application Support/Claude, and ~/.continue.

›Can I perform an enhanced scan of my MCP servers?

Yes, you can provide an Anthropic Claude API key using the --claude-api-key flag to enable deeper vulnerability analysis.

›How can I exclude specific servers from the security scan?

You can use the --safe-list flag followed by a comma-separated list of server names to skip them during the scanning and cross-origin violation detection process.

›Why would I use the --identify-as flag?

The --identify-as flag allows you to test how MCP servers behave when identifying as a specific client, such as claude-desktop, which helps detect potential bait-and-switch security issues.

›Does MCP-Shield support scanning custom configuration files?

Yes, you can point the tool to a specific configuration file by using the --path flag followed by the file path.

Badge

Maintain this server? Add the live badge to your README.