@forkpoint/agent-lighthouse-mcp
Model Context Protocol server for Agent Lighthouse. It lets Claude Desktop, Cursor and other MCP clients scan a website for AI-agent readiness.
Setup
{
"mcpServers": {
"agent-lighthouse": {
"command": "npx",
"args": ["-y", "@forkpoint/agent-lighthouse-mcp"]
}
}
}
The package's command is agent-lighthouse-mcp (dist/bin.js). It speaks MCP over stdio. Logs go to stderr; set LOG_LEVEL and LOG_FORMAT to change them.
The audit_website tool
| Argument | Required | Meaning |
|---|---|---|
url | yes | The URL to scan. |
pageType | no | homepage, category, product, article or general. Omit to detect. |
pages | no | More { url, pageType } declarations. |
snapshots | no | Browser-rendered DOM input. |
It returns JSON: overall score and tier (null with unscoredReason when the scan was not scored), per-category scores, readiness vitals, page scope, and the top 10 fixes. Progress is sent as notifications/progress when the client passes a progress token.
auditWebsite(url, { pageType, pages, snapshots }), exported from this package, runs the same scan from code.
The registry_summary tool
Takes no arguments and runs no scan. It returns the installed version's getRegistrySummary() as JSON: the core version, the number of audits and categories, the count per tier, the registry mass, and the same per category.
Links
- Documentation: https://forkpoint.github.io/agent-lighthouse/
- Repository: https://github.com/ForkPoint/agent-lighthouse
- CLI package: https://www.npmjs.com/package/@forkpoint/agent-lighthouse
License
Apache-2.0