AWS launches QA Studio: Natural language test automation powered by Amazon Nova Act

AWS launches QA Studio: Natural language test automation powered by Amazon Nova Act

AWS has released QA Studio, a reference solution for test automation built on Amazon Nova Act that replaces code-dependent testing frameworks with natural language specifications and visual understanding.

What QA Studio Does

QA Studio is an open-source reference implementation deployed through AWS CloudFormation that enables teams to define quality assurance tests in natural language. Instead of writing code that relies on brittle UI selectors and element identifiers, teams describe test steps the way product managers define acceptance criteria.

The system is powered by Amazon Nova Act, AWS's custom computer use model that interacts with applications through visual understanding and natural language instructions—the same way users do. When developers refactor code or designers adjust layouts, tests automatically adapt rather than breaking.

Key Capabilities

Natural Language Test Management: Teams define test suites using the same language used to describe product requirements. The "User Journey Wizard" generates test cases from user journey descriptions, and Amazon Nova Act translates these into browser actions without requiring code authorship.

Visual Navigation: Rather than relying on code-dependent selectors and DOM inspection, Nova Act navigates applications based on visual appearance and context. This removes the brittleness that forces teams to maintain tests whenever UI implementation changes.

End-to-End Visibility: The system captures trajectory logs showing what the agent saw and why it took specific actions at each step. Test recordings, screenshots, and reasoning traces are surfaced in QA Studio's interface, enabling debugging through natural language context rather than parsing stack traces.

Scalable Execution: QA Studio runs on serverless AWS infrastructure including Lambda, ECS with Fargate, SQS, and EventBridge. Tests can execute on-demand, on schedules, or integrated into CI/CD pipelines.

Architecture

QA Studio is built on AWS services including:

• Amazon Nova Act for agentic UI automation
• Amazon Bedrock for foundational model access and managed remote browsers
• AWS Lambda and ECS/Fargate for backend processing and test execution
• Amazon DynamoDB for test definitions and execution history
• Amazon S3 for test recordings and logs
• Amazon SQS for reliable queue-based test processing
• Amazon EventBridge for test scheduling
• Amazon Cognito for authentication

The architecture provides automatic scaling with consumption-based pricing across all services. Teams deploy QA Studio in their own AWS account using AWS CDK, maintaining complete control over security policies, compliance requirements, and test data isolation within their security boundary.

Availability and Deployment

QA Studio is available as an open-source GitHub repository. Teams clone the repository and deploy infrastructure using AWS CDK templates included in the project. The README provides step-by-step deployment guidance, optional CI/CD integration configuration, and resource cleanup instructions.

All test data, recordings, logs, and traces remain within the customer's AWS account. Organizations can configure VPC settings and access controls according to their requirements.

What This Means

This addresses a fundamental pain point in QA automation: the maintenance burden created by tight coupling between test code and UI implementation details. By using visual understanding instead of code inspection, QA Studio reduces the distance between how teams describe requirements and how tests validate them.

The approach democratizes test authorship—teams no longer need specialized automation engineering knowledge to maintain test suites. However, the practical impact depends on Nova Act's consistency in visual navigation and its handling of complex or dynamically-rendered interfaces. Early adoption will reveal whether vision-based automation can reliably replace selector-based frameworks at scale, or whether hybrid approaches will emerge as more practical for production deployments.