What does MCP Defender do?

MCP Defender is a protocol that allows AI models to interact with tools and systems. It protects AI applications from malicious tool calls and attacks.

What does MCP Defender scan?

Currently, MCP Defender scans all MCP tool call traffic, both requests and responses. In the near future, the tool descriptions will also be scanned.

Can I disable protection for a given tool?

Yes, you can choose to always allow a certain MCP tool even if MCP Defender detects something malicious.

Is MCP Defender free to use?

Yes, MCP Defender is free to use. You can use your MCP Defender account which will give you 250 scans per month. Upgrading to Pro will give you 5000 scans per month. Alternatively, you can use your own API keys.

What types of MCP servers do you support?

Currently we support STDIO servers but you can also protect SSE servers using mcp-remote. SSE and Streaming MCP Server support are coming soon.

What data do you collect?

Currently, we do not collect any data if you use your own LLM provider’s API key. If you choose to use an MCP Defender account to scan instead, then we will store your scan requests and results in our logs. These logs will not be shared with anyone and will only be accessed to debug issues. We will eventually have a way to run it in private mode to disable log collection on our backend.

MCP Defender : Reviews, Pricing, Alternatives & Features (2026)

Newsletters.ai — Learn about AI, the lazy way.

Available Via:Desktop

Visit MCP Defender

MCP Defender is a desktop app for individuals running AI coding and chat tools that need real-time security for Model Context Protocol (MCP) traffic between AI apps and MCP servers.

It works as a secure proxy that intercepts MCP tool call requests and responses, then checks them for malicious activity before it reaches the AI app or connected tools:

Real-time threat detection using multiple layers, including LLM-powered analysis and deterministic security signatures
User alerts and approval workflows to allow or block suspicious tool activity
Coverage for common attack classes such as prompt injection, credential theft, arbitrary code execution, and remote command injection
Scan controls to manage signatures used during scanning, plus the option to choose an LLM provider and use personal API keys

Pricing: Undisclosed, contact for pricing.

Notable for being open source and AGPL-3.0 licensed, and for targeting MCP-specific risks by scanning tool calls in supported apps such as Cursor, Claude, Visual Studio Code, and Windsurf.