Open-source AI firewall that proxies MCP tool calls to detect and block prompt injection and other threats.
What it does
MCP Defender is an open-source (AGPL-3.0) security tool that acts as a proxy between AI applications and MCP servers, scanning all MCP tool calls in real time. It combines LLM analysis with deterministic signatures to catch prompt injection, credential theft, and code or command injection. It has been acquired by Docker Inc.
Core features
AI firewall proxy for MCP tool calls
LLM-powered threat detection
Signature-based scanning controls
Bring your own LLM provider and API key
Supports Cursor, Claude, VS Code, and Windsurf
Open source, AGPL-3.0 licensed
Best for
→Protecting AI coding assistants from prompt injection
→Preventing credential theft via MCP
→Blocking malicious tool calls