Cuahttps://cua.ai/

Your landing page’s weakest score is Mobile web.

How Cua ranks

  1. 1.E2B80
  2. 2.Kernel77
  3. 3.Browserbase76
  4. 4.Cua64
  5. 5.Morph55
  6. 6.Scrapybara48
Fix Mobile web
64Growth scoreupdated

category scores

Mobile web54Information hierarchy62
category scoresShowHide
Mobile web54Information hierarchy62

Mobile web

54Fair

The phone experience as a first-class citizen.

Mobile web annotations on the mobile capture123
  1. Hero animation panel — Decorative ASCII-art frame consumes the top half of the fold and pushes the headline and any CTA out of view.
  2. Truncated announcement bar — Banner cuts off mid-sentence ('6 of 25 expert Ki…') with no visible link, leaving a broken message at the top.
  3. /ask input field — The only actionable element sits below the fold and reads as a static code snippet, not a tappable CTA.

Information hierarchy

62Fair

Does the design help or inhibit what the page is trying to say?

Information hierarchy annotations on the desktop capture1234
  1. Fleets nav button — The only button-styled CTA is a small corner nav item with a vague noun label, invisible in a one-second scan.
  2. /ask input field — Competes with the nav button as the apparent primary action; the fold never signals which one to take.
  3. Navigation link row — Ten-plus links plus a Discord icon outweigh the CTA and leak attention before any conversion.
  4. Cua-Bench banner link — 'VIEW OVERVIEW' pulls first-glance attention to an exit path at the very top of the page.

CTA

62Fair

Copy, size, verb, cost-feel — does the primary CTA earn the click?

Cuain-product action

How Cua compares

Recommended rewrites

Trust

64Fair

Proof that the product is real, competent, and safe to try.

Cua desktop capture for Trust

How Cua compares

Trust read

Value props

76Fair

Title, subtitle, hero — do they sell the product's true value?

Cua’s headlineScale computer fleets for every agentRun computer-use agent training, eval, and data-generation workloads in parallel on Linux, Windows, macOS, and Android machines. Fork from snapshots, reproduce failures, and turn agent activity into training data.

Working

Could improve