TLDR: Amazon Web Services (AWS) has released a preview of its Nova Act SDK, designed to enable developers to build and deploy robust browser automation agents. This SDK leverages a new AI model, Amazon Nova Act, specifically trained for understanding and interacting with web interfaces, aiming to overcome the limitations of traditional rules-based automation and enhance productivity for browser-based workflows.
Amazon Web Services (AWS) has announced the preview release of its Nova Act SDK, marking a significant step towards more reliable and scalable browser automation through artificial intelligence. Initially introduced as a research preview in early 2025, the Nova Act SDK is now moving into a limited preview, allowing customers to transition their browser automation agents into production environments.
This SDK is powered by the Amazon Nova Act model, an AI specifically purpose-built and fine-tuned for ‘agentic AI’ – systems capable of performing actions within a web browser. The model is trained using reinforcement learning and extensive in-domain browser interaction data, enabling it to execute step-by-step workflows with precision. This approach aims to address the inherent brittleness of traditional rules-based browser automation frameworks, which often break due to minor changes in web page structures.
The Nova Act SDK integrates seamlessly with existing AWS services, enhancing its production readiness. These integrations include AWS Identity and Access Management (IAM) for secure credentialing, Amazon Simple Storage Service (Amazon S3) for data storage and policy control, and the new Amazon Bedrock AgentCore Browser Tool for scalable, cloud-based browser execution. This ensures that agents built with Nova Act can operate within enterprise environments with the necessary reliability, observability, and security.
Developers can define agents using Python, natural language, or a combination of both, facilitating rapid prototyping and iteration. The SDK supports features such as Git-based CI/CD pipelines, reusable agent modules, Playwright fallbacks for sensitive tasks like password input, and threading/parallelization for faster execution. This flexibility allows teams to move from prototype to production in days, rather than weeks.
Amazon envisions agents as systems that can complete complex, multi-step tasks across various digital and physical environments, moving beyond simple natural language responses or knowledge base retrieval. The Nova Act SDK is designed to break down complex workflows into reliable atomic commands (e.g., search, checkout, answer questions about the screen) and allows for detailed instructions to be added. It also supports calling APIs and direct browser manipulation through Playwright for enhanced reliability.
Potential use cases span various industries, including automating back-end e-commerce tasks, managing vendor portals, processing invoices, accessing internal dashboards, and filling out repetitive forms. The SDK is particularly valuable for workflows that reside entirely within web browsers and lack robust API coverage. While Amazon acknowledges that Nova Act is a research preview and not a finished product, early benchmarks suggest competitive performance against other AI models in browser interaction tests.
Also Read:
- Accelerating Drug Discovery: Building AI Research Assistants with Strands Agents and Amazon Bedrock
- CyberArk Enhances AI Security with New Tools in AWS Marketplace
The release of the Nova Act SDK underscores AWS’s commitment to advancing AI-driven automation, inviting developers to explore new possibilities and build more robust and adaptable browser agents.
