Norveth
← Back to Docs
Integration · Model Context Protocol (MCP)

Norveth inside your AI assistant.

Norveth speaks the Model Context Protocol, so any MCP client — Claude, ChatGPT, Cursor, Windsurf, Cline — can analyze a website URL and ask grounded questions about its performance, UI/UX, and security. One API key, eleven tools.

Availability

The hosted endpoint at /api/mcp is gated until enabled on this deployment. The stdio package @norveth/mcp is built for private use and is not published to npm yet — contact us if you need access.

01

Get your API key

Sign in at norveth.app (the free plan works), open Settings → API Keys, and create a key starting with norveth_sk_.
02

Claude Desktop, Cursor, Windsurf & Cline

These run the server locally over stdio. Add this to your client's MCP configuration:

{
  "mcpServers": {
    "norveth": {
      "command": "npx",
      "args": ["-y", "@norveth/mcp"],
      "env": { "NORVETH_API_KEY": "norveth_sk_your_key_here" }
    }
  }
}

Restart your client, then ask: “Analyze https://example.com with Norveth and tell me its biggest security and performance issues.”

03

ChatGPT & Claude.ai Web Connectors

Web clients connect to Norveth's hosted MCP endpoint directly over Streamable HTTP — nothing to install locally:

https://norveth.app/api/mcp

Authenticate with your norveth_sk_ key as a Bearer token.

Available MCP tools (11)

analyze_repositoryStart a 3-dimensional analysis of any website URL.
get_analysis_statusPoll an analysis job until it's COMPLETED.
list_projectsList or search website scans in your organization.
get_engineering_reportSummary + performance, UI/UX, and security scores.
get_security_findingsSecurity header policies, TLS cert health, and OWASP findings.
search_codebaseSearch the site's crawled content and structure.
ask_about_projectAsk anything; get an answer grounded in deterministic website audit data.
list_pagesList the crawled pages captured for an analyzed website.
get_page_sourceFetch a page's full source — raw HTML, text, and asset inventory.
get_source_treeRead a site's reconstructed source (original files from source maps + bundles).
get_site_intelligenceThe single JSON with everything: overview, tech, probes, pages, source, findings.