2.0.0 - 75% cheaper test costs, dashboard, UX improvements and more!

ATP360 2.0 focuses on three things: more accurate tests at a lower cost per run, clearer visibility into what your tests are doing and what they're costing, and a faster path from deployment to your first test. [[embed: https://www.youtube.com/watch?v=KU4TE87L7ng&feature=youtu.be]](https://www.youtube.com/watch?v=KU4TE87L7ng&feature=youtu.be) AI engines and accuracy The latest models, out of the box ATP360 2.0 supports Opus 5, Gemini 3.1 and GPT 5.6 with no extra configuration. Each stage of the AI engine pipeline (parser, orchestrator and execution) can use a different provider and model, and you can change the selection immediately before an individual test run without altering your workspace defaults. Change engines for a single run, or set workspace defaults in Settings. More precise GUI interaction The CUA interacts with the GUI more precisely in 2.0. It spends less time closing things that shouldn't have been opened, and fewer attempts repeating a step that failed first time around. The result is faster tests and lower AI costs on the same work. Full 1920 x 1080 support When an area of the screen is too small to read reliably, the CUA now zooms into the screenshot for a closer look. This brings support for 1920 x 1080 resolution for the first time. Visibility and control A new home dashboard Sign in to see the state of your testing at a glance: tests awaiting review, total tests run, pass rate, total running cost and average cost per test. Below the headline figures are trends for pass and fail volumes, failures by type, override rate, execution by OS and how much of your output has been reviewed. Charts are interactive, so you can move from any data point straight to the tests behind it. Headline numbers and trends, with click-through to the underlying tests. New limits and controls A new Limits & Controls tab in Settings & Administration sets financial, usage and reliability boundaries for AI execution, so no run can loop away unnoticed. Configure maximum actions per test step, turns per run, reprompts per task, session restarts, a maximum cost per test run, and how frequently the orchestrator gates itself. Tune them against your own test estate to trade cost against reliability, and return to the shipped defaults at any point. Orchestrator controls, with a one-click return to defaults. Clearer messages when a test fails Failures now lead with a plain-language reason and keep the technical detail collapsed until you need it. Where a failure type is documented, the message links straight to the relevant guidance, making it quicker to resolve yourself or to give support what it needs first time. A readable failure reason, with the underlying detail one click away. Working with a live test A live test view you can arrange to suit you Resize the test steps pane and the actions log independently, or collapse them altogether to give the screen over to the live view. Live view and screenshots can now be expanded to full screen for a closer look at what is happening on the machine. Steps are ticked off in real time as they complete, and each action is correlated to the step that triggered it, so the run is auditable after the fact. Success and fail criteria are visible throughout the test rather than only at the end. Test steps, live view and action log side by side. Both panes collapsed, giving the full width to the live view. Steps ticked off as they complete, with success and fail criteria visible during the run. Deployment and administration A simple setup wizard Once you have deployed the ATP360 offering from the Azure Marketplace, the new setup wizard takes you through connecting your AI engines and pairing your CUA workers. Configuration that previously needed several trips through Settings is now a single guided pass, so you can get to your first test in minutes. An enterprise-ready CUA worker The CUA worker now ships as a signed MSIX package, so it deploys cleanly through Intune, Configuration Manager or whichever tooling you already run. It carries fewer third-party dependencies, which means less for your team to patch and maintain. Connectivity between worker and orchestrator has also been reworked. Connections are now initiated by the worker, which removes the need for inbound access to your test machines. Network roaming is handled more gracefully, and workers recover reliably after an administrator has connected over RDP between tests. Cleaner CUA worker management The CUA Workers view now shows OS version and agent version alongside health for every worker, making it obvious which machines are offline or running an older agent. Add, edit and remove workers from the same place. Health, OS and agent version for the whole fleet in one view.