What it does

AI agents increasingly browse websites and complete tasks on them, such as booking an appointment or placing an order. The Agentic browsing category in Lighthouse audits a website for accessibility, visual stability and WebMCP integration. WebMCP is a proposed standard for exposing tools to AI agents on existing websites. Modern Web Guidance includes a webmcp skill that lets a developer delegate WebMCP tool implementation to a coding agent. Chrome DevTools for Agents turns an AI-assisted IDE or coding agent into a browsing agent. That agent can simulate the steps an agent would take. It also provides logging and screencasts that show how the agent perceives and interacts with the page. Agentic browsing is informational and unbenchmarked, so it reports pass or fail status and warnings rather than a ranking.

Capabilities

Lighthouse Agentic browsing audits are available starting from M150.
The accessibility audit verifies a subset of the Accessibility audit categories, including that every interactive element has a programmatic name.
The stability audit measures Cumulative Layout Shift to flag elements that move unexpectedly and could cause agent misclicks.
The WebMCP audit checks for registered WebMCP tools, forms missing declarative WebMCP, and schema validity.
Chrome DevTools for Agents can invoke the lighthouse_audit tool directly on the active tab for a multi-category health check of the page's current state.

When to use it

You want to assess how agent-friendly your website is.
You want to check the machine-readable signals an AI agent needs for a flow on your website, such as booking an appointment or placing an order.
You want to reproduce agent failures and check that your website's flows are deterministic.