MCP Server for AI Agents

Add npmscan's package and vulnerability lookups to your AI agent or skill in one line — no API key, no scraping.

Overview

NPMScan runs a remote MCP (Model Context Protocol) server at https://npmscan.com/api/mcp. MCP is an open standard, not Claude-specific — Claude Code, Claude Desktop, ChatGPT (via Developer Mode), Cursor, and other MCP-capable clients can all connect to it. One URL gives your agent 23tools for npm package and vulnerability lookups, supply-chain risk signals, license compliance, and remediation ranking. No API key required; it's the same public, read-only data as the website.

npmscan is now an approved app in the ChatGPT Apps directory — install it there for one-click setup instead of adding the connector manually.

Claude Code users can also install the npmscan Claude Code plugin, which wraps this MCP server with two bundled skills — dependency-audit and package-trust-check — for whole-tree scans and single-package trust investigations. See the repo for install instructions.

Quick Setup

Claude Code

claude mcp add --transport http npmscan https://npmscan.com/api/mcp

Claude Code project/user config (.mcp.json or ~/.claude.json) and other clients that accept an explicit type

{
  "mcpServers": {
    "npmscan": {
      "type": "http",
      "url": "https://npmscan.com/api/mcp"
    }
  }
}

Stdio-only clients (bridge via mcp-remote) — also the safe fallback for Claude Desktop

{
  "mcpServers": {
    "npmscan": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://npmscan.com/api/mcp"]
    }
  }
}

Claude Desktop: Settings → Connectors → Add custom connector, then paste the URL. ChatGPT: Settings → Connectors → Advanced → Developer Mode (Plus/Pro/Business/Enterprise/Edu only), then Add custom connector.

Available Tools

The server exposes 23 tools, from basic package/version lookups to install-script analysis, publish-provenance and maintainer-change checks, license compliance, and remediation ranking. Every result includes an npmscanUrl field linking back to the full write-up on npmscan.com.

See every tool's input schema and real request/response examples →

Limits & Notes

No authentication — public, read-only data.
Rate-limited to 30 requests/minute per IP on this endpoint.
Stateless Streamable HTTP transport; SSE is disabled (it's deprecated in the current MCP spec).
Upstream data comes from the npm registry, OSV.dev, GitHub Advisories, NIST NVD, MITRE, CISA KEV, and FIRST.org EPSS — subject to their own availability and rate limits.