MCP Radar

MCP servers are becoming a new supply chain for AI agents. Some are great. Some are abandoned. Some are openly dangerous.

MCP Radar is a public registry that scores servers on permissions, maintenance, install method, package quality, docs, and security risk — so you can evaluate one before plugging it into Claude Desktop or Cursor.

The model is open. Every score links to the signals that produced it.

• Who maintains this MCP server?
• What permissions does it require?
• Is the package pinned and released responsibly?
• Are there known vulnerabilities?
• How does it compare to alternatives in the same category?

One command. No account. No telemetry.

1# install
2npm install -g @agentopssec/mcp-radar
3 
4# look up a server by repo
5mcp-radar score modelcontext/github-mcp
6 
7# compare two alternatives
8mcp-radar compare github-mcp filesystem-mcp
9 
10# grab a badge for your README
11mcp-radar badge owner/repo

Identical category, very different signals.

Public, transparent, consumable by local tools.

1{ "name": "example-mcp-server",
2  "score": "B-", "risk": "medium",
3  "categories": ["github", "filesystem"],
4  "findings": [
5    { "type": "permission", "severity": "medium", "message": "Requires repository write access" },
6    { "type": "maintenance", "severity": "low", "message": "Last release was 22 days ago" }
7  ] }