MCP server risk checker.

Before you connect an MCP server to an AI assistant, check it. Fourteen questions, a risk rating and a fix list.

Runs in your browser. Your answers stay on this page. Nothing is sent or stored.

Check a server Why it matters

Checker

Should you trust this MCP server?

Answer for the server you are about to install or connect. Choose Unsure if you do not know.

Why it matters

An MCP server acts with your access.

The risk

It can do what you can.

An MCP server gives an AI assistant tools: reading files, calling services, running commands. A careless or malicious one can leak data or take actions you never intended.

Hidden instructions

Tool descriptions are read by the AI.

Text in a tool description or a tool result can steer the model. This is why unreviewed servers and untrusted content are a prompt-injection risk.

The fix

Least access, with a human in the loop.

Trusted source, pinned version, narrow permissions, a sandbox, and approval before anything is changed or sent.

A checklist for review, not a security test. For a thorough review, AI security assurance ›

Free live toolCloud or local AI chooser Free live toolData classification helper Free live toolPhishCheck MCP