AgentArmorHQ
Sign in
Add-on review

Which AI add-ons can your engineers use?

Paste the names of the MCP servers (add-ons that connect your AI to other tools) and skills your team uses. Most answers take under two minutes, a few up to five. You see which to allow and which to block, and if we flag something, the file and line we flagged.

  1. 1Paste namesOne per line, or a whole config file.
  2. 2We read each add-on's public codeWe need nothing from you.
  3. 3You get an answer and a list to pasteFor Claude Code and Cursor.

One per line, up to 10: an npm package, pypi:name, or a GitHub link. Or paste your MCP config file (.mcp.json, claude_desktop_config.json, ~/.cursor/mcp.json). From a config file we take only each package name, or the host of a hosted server. The rest of the file stays in your browser.

Only if you want a copy of the results. You do not need it to see them.

Free for up to 10 add-ons. No account, no card, no email needed. We never need access to your code.

Not sure what your team uses? Ask an engineer for their MCP config file (.mcp.json or ~/.cursor/mcp.json) and paste it here.

What we catch and what we don't

Block: a public security report names this add-on or version.

Look closer: our reading of the code found something worth a person's look.

We don't promise to catch brand-new malware that nobody has reported yet. In our test, code reading flagged about half of it.

A check lowers risk. It does not remove it. We say "Block" only when a public security report names the add-on or version. When we are not sure, we say "Look closer".