AgentX
Agent Experience analytics

See your site the way AI agents do.

More of your visitors are agents buying, booking and researching on behalf of people. AgentX shows whether they can find what they came for, and lets them tell you what got in the way.

Free Agent Readiness audit

Checks robots access for AI agents, llms.txt, MCP discovery, structured data, labelled controls and token weight. Takes a few seconds.

Try the demo store: this server's /demo/

Add your site

Free. You get a site ID, a 1-line snippet and a dashboard.

What the snippet does

  1. Advertises a feedback channel to agents with <link rel="agent-feedback">, a manifest and plain-language instructions in the DOM.
  2. Gives you an MCP endpoint so MCP-native agents can file feedback as a tool call.
  3. Exposes window.__AX_FEEDBACK__() for browser agents (Playwright, computer-use) to report each task.
  4. Feeds your dashboard: success rate by intent, friction causes, failing selectors, agent families and how you compare to your industry.

How agents report feedback

One report per task attempt. Full schema at /v1/protocol.

await window.__AX_FEEDBACK__({
  taskIntent: "Buy a black t-shirt, size M, under $40",
  taskSucceeded: false,
  navigabilityScore: 2,
  frictionCategory: "unlabeled_controls",
  selectorFailed: "button.icon-cart",
  agentNotes: "Size picker buttons have no labels; could not tell M from L.",
  agentFamily: "claude", agentModel: "claude-opus-5-5"
});

Network benchmarks

Across all sites reporting in the last 30 days.

Loading...