Synthetic monitoring simulates real user journeys to find problems early. Instead of waiting for customers to report issues, teams run automated tests regularly.
Traditional methods rely on browser automation tools like Selenium, which depend on CSS selectors. Small UI changes often break these tests, requiring constant updates.
Amazon Nova Act uses a large language model that reasons from UI screenshots. This makes it more adaptable to style changes and reduces maintenance. Early tests show over 90% accuracy in browser workflows.
The architecture combines several AWS services. Amazon Nova Act defines and runs UI workflows using natural-language actions. Amazon Bedrock AgentCore hosts the agent in a serverless environment. The AgentCore Browser provides a secure, remote browser for each test.
Scheduled tests run every few minutes, triggered by Amazon EventBridge Scheduler. If a journey fails, the agent sends a notification through Amazon SNS. This setup offers fully managed, resilient synthetic monitoring without brittle scripts.
To use this, teams need access to Amazon Nova Act, Amazon Bedrock AgentCore, Amazon ECR, IAM, EventBridge Scheduler, and SNS. Basic knowledge of Python, Docker, and AWS CLI is also required.