AgentScore MCP server

This is a small helper that lets your AI answer questions about the safety of software packages, especially ones built for AI assistants themselves. It scans those packages for risky behaviour and tells you how much to trust them. If you ever wonder "should I really install this thing someone mentioned online", this is for you.

What this MCP server does

Once connected, you just ask your AI something normal like whether some package is safe. Your AI sends that question to this helper behind the scenes. The helper then runs a real security scan on the package, checking things like hidden install steps, suspicious links, leaked passwords inside the code, and who published it. A moment later your AI replies with a clear score, a simple risk level, and the reasons why. You never leave the chat window.

Click to zoom

What you can do with it

• Get a yes-or-no style trust verdict on almost any AI-related package
• See a numbered safety score plus a low-to-critical risk label
• Find out what warning signs were spotted during a scan
• Discover which popular packages have recently been hacked or flagged
• Look across all the AI helpers already listed in your own project folder
• Generate and drop in an automated checker that reviews each update request

Try asking your AI

• “Is this package called mcp-filesystem-safe okay to install? What did the scanner flag on my last installed plugin? Which well-known AI plugins rely on the library everyone says got hacked today? Please put the automatic safety-checking job into our team repository.”

What it gives back to you

Mostly short readable answers right in the conversation. For a single package you typically get a number out of 100, a word such as moderate or high, and a couple of lines naming whatever looked off. Asking about many items returns a compact summary listing each one next to its result. Setting up automation writes a configuration file straight into your working directory rather than printing long output.

Before you start

Who it's for

Anyone adding third-party extensions to their AI toolkit, particularly developers and IT staff responsible for keeping company projects clean and secure.