August 2026 release notes for Test Manager in Automation Cloud and Test Cloud.

Performance Testing adds a dedicated Reports tab for each scenario. From this tab, you can:

For details, see Analyzing performance results.

Dry-run performance reports include probe timeline data in all production environments, following the same rollout already completed in lower environments. For details, see Analyzing performance results.

The manual execution assistant shows the test case description in a collapsible panel above Precondition, so testers have the full context without leaving the execution window. For details, see Executing manual tests.

Exporting requirements and test cases to Excel is asynchronous. You are notified when the file is ready instead of waiting on the request. Both exports also gain one column per custom field definition, instead of a single column concatenating all values, and test case exports let you collapse step rows to keep large exports scannable. For details, see Export data.

Test Manager consistently shows whether a test case already has an automation linked everywhere you view test cases, not only from within a test set. For details, see Selecting automation.

Performance Testing

Exports

Test cases and requirements

Migrations

Integrations and accessibility

Performance Testing now scans your automation before a run starts and flags anything that will not work at scale, including unsupported activities (computer vision, image-based/OCR, Hardware Events input mode, local file system access, and desktop-only interaction patterns) and Studio version or activity-package mismatches. This surfaces problems before you spend robot minutes on a run that was never going to succeed. For details, see Creating automated tests.

Instead of estimating virtual-user capacity from static CPU/RAM ratios, Capacity Planner runs a short series of safe load probes against your actual hardware and extrapolates a data-backed virtual-user recommendation in minutes. Built-in guardrails keep the recommendation within limits already proven safe on your infrastructure. For details, see On-premises robot system requirements.

If port 5671 (AMQP to Azure Event Hub) is blocked, Performance Testing automatically falls back to WebSocket-based communication over port 443, and if that also fails, to API-based metric delivery. You no longer need to request a new firewall rule to get metrics flowing. For details, see Software and system requirements and Troubleshooting performance testing.

Robot logs are now viewable and filterable, by load group and by severity, directly inside the Performance Testing results screen, alongside application logs. For details, see Analyzing performance results.

Elastic Robot Pools generate load directly from your own Azure, AWS, or Google Cloud private capacity instead of shared, multi-tenant infrastructure, scaling load injectors up and down as a test requires. For details, see Platform Units and infrastructure costs.

Every performance test run now generates a dedicated report view for each stakeholder who needs to act on the result: Program Managers get a ship-it-or-not verdict, Developers get what is broken, InfraOps gets infrastructure issues, and QA/Testers get what needs to be retested. For details, see Analyzing performance results.

We are deprecating the GPT-4o model in Autopilot. Support for GPT-4o will end on October 31, 2026. Before this date, switching to a supported alternative model in your AI Trust Layer Policy avoids disruption to test generation, quality checks, and other Autopilot capabilities that rely on model selection.