Track Your Lighthouse Agentic Browsing Score
See how ready your site is for AI agents, tracked automatically over time.
What it checks
Lighthouse 13.3 added an Agentic Browsing category with six equally-weighted audits:
| Audit | What it checks |
|---|---|
| WebMCP tools registered | Tools registered on document.modelContext at scan time |
| WebMCP schemas are valid | Each tool's inputSchema is well-formed |
| WebMCP form coverage | Forms on the page missing a WebMCP annotation |
| Accessibility tree is well-formed | Whether AI agents can navigate the page structure |
| Cumulative Layout Shift | The standard CLS metric |
| llms.txt follows recommendations | Presence and format of your llms.txt file |
The three WebMCP-specific audits rely on a Chrome DevTools Protocol domain that only exists in Chrome builds shipping the native WebMCP implementation (Chrome 149+, matching the origin trial). On an older Chrome, those three show as not applicable rather than pass or fail. Accessibility tree, CLS, and llms.txt work on any modern Chromium.
Where to see it
Every site's dashboard page has an Agentic Browsing Score panel showing your latest score, a trend across recent scans, and a pass/fail/not-applicable breakdown for each audit with a link to learn more.
How scans run
Scans run as a background job, not on page load, so viewing your dashboard never triggers a live Lighthouse run. Each site gets rescanned roughly once a week.
Want a fresh score right after a change? Click Scan now next to the panel to trigger one on demand. It's capped at one scan per site every 5 minutes, and queues behind any scan already in progress, so it can't be used to run Chrome concurrently against your app.
If you're self-hosting Aigentably, wire up the scheduled scan with cron:
0 3 * * * cd /path/to/webmcp-hub && npm run lighthouse-scanThe script only scans sites that haven't been checked in the last 7 days, and runs one Chrome instance at a time so it doesn't compete with your app for CPU or memory. Point it at a specific Chrome/Chromium binary with the LIGHTHOUSE_CHROME_PATH environment variable if the default isn't picked up.
If a scan fails
If Aigentably can't reach your site (it's down, DNS is misconfigured, or it's behind auth), the panel shows the error message from the last attempt instead of a score. The next scheduled scan will retry automatically.