Webskan
Audit & re-check the website your AI agent shipped
Start new thread

Webskan - Audit & re-check the website your AI agent shipped

by•
Remote MCP server that lets Claude Code, Cursor, Codex or VS Code audit a live site: Lighthouse, technical SEO, accessibility, security headers, broken links across 30 pages. Stable finding ids and a recheck diff: fixed / open / new. Free plan, no account.

Add a comment

Replies

Best
Maker
📌

Hi Product Hunt! Webskan started from a simple observation: coding agents are great at shipping pages and bad at noticing what they broke two pages away. The build is green, the deploy succeeded, and the site has placeholder titles, a leftover noindex, broken links in the footer, no security headers and a readable .env in the web root.

Webskan gives the agent the outside view. It is a remote MCP server (Streamable HTTP, nothing to install) with four tools: scan_website (30-second Lighthouse check), audit_website (whole site, up to 30 pages: technical SEO, axe-core accessibility, security headers and TLS, publicly exposed .git/.env/SQL dumps/backups, broken links, forms and key journeys), get_audit and recheck_website, which returns a diff: fixed / still open / new. Finding ids are stable, so the agent can prove what it fixed.

What is free: all four tools, right now, without an account. 2 whole-site audits per IP per day, quick scans 10 per hour.

What is paid: a Developer plan at $29/month for 30 audits a day per key and no one-audit-per-host-per-day window, so you can re-audit five minutes after the fix. Next on the list: a GitHub Action for the deploy pipeline.

Caveats, honestly: public URLs only, Lighthouse scores vary a few points between runs, and a full audit takes 1 to 5 minutes. I would love to hear which findings are noise for you and what is missing. Docs: