Defender MCP Server
Proxy and inspect MCP tool traffic to detect potential security threats automatically.
Data last scanned today · Reviewed today
Overview
MCP Defender acts as a security proxy for AI applications, monitoring all MCP tool call requests and responses. It intercepts traffic between your AI agent and tools, checking the data against predefined security signatures to identify potentially malicious activity. When a threat is identified, the application provides an interactive prompt, allowing you to manually choose whether to block or allow the suspicious tool call. It supports integration with popular AI environments including Cursor, Claude, Visual Studio Code, and Windsurf.
Our verdict
This project currently holds a security score of 34/100, which results in a high-risk classification. The score is primarily impacted by a lack of editorial review, unverified authentication methods, and the fact that it is a community-maintained project rather than an official vendor release. Given these factors, users should exercise caution. It is suitable for those looking to monitor local MCP traffic, but it should not be relied upon as a hardened security solution until further reviews are conducted.
- Supports active protection for Cursor, Claude, VS Code, and Windsurf.
- Intercepts and proxies all MCP tool traffic locally.
- Provides manual approval prompts for blocked tool calls.
- Maintains a relatively small footprint with 37 direct dependencies.
- Security score of 34/100 indicates high risk level.
- Authentication methods have not yet been reviewed.
- Read-only capabilities remain unverified.
- Last commit was 95 days ago.
Setup
{
"mcpServers": {
"mcp-defender": {
"args": [
"-y",
"mcp-defender"
],
"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
›Which AI applications are supported by MCP Defender?
The server automatically protects Cursor, Claude, Visual Studio Code, and Windsurf.
›How does the software handle potentially harmful tool calls?
When the application detects a threat, it alerts the user and prompts them to either allow or block the specific tool call.
›Does this tool run as a standalone application?
Yes, MCP Defender is a desktop application that intercepts and proxies MCP traffic to analyze data against security signatures.
›Are there options for installing the software other than the binary download?
You can clone the repository, install the necessary dependencies using npm install, and launch the application with npm start.
Changelog
Badge
Maintain this server? Add the live badge to your README.