Build reliable AI agents for UI workflow automation with Amazon Nova Act, now generally available - Amazon Web Services

Earlier this year, we released a research preview of Nova Act, demonstrating the potential of AI agents to interact with user interfaces and automate complex workflows. Developers experimented with Nova Act and told us they wanted to bring these automation agents to production.

But bringing agents to production required much more than just model access. Developers were spending significant time orchestrating workflows, refining prompts, choosing the right tools, and stitching together disparate components to achieve reliable automation. The challenge wasn’t just intelligence—it was reliability, integration, and speed to production. So we built a fully integrated solution for production-ready browser automation.

Today, we’re announcing the general availability ofAmazon Nova Act, a newAmazon Web Services (AWS)service that helps developers build, deploy, and manage fleets of reliable AI agents for automating production UI workflows. Nova Act delivers over 90% task reliability at scale while offering the fastest time to value and ease of implementation compared to other AI frameworks.

Here’s a quick look at the Nova Act console.

Nova Act addresses the challenge of building reliable browser automation at enterprise scale. Powered by a custom Amazon Nova 2 Lite model, Nova Act excels at driving browsers, support calling APIs, and escalating to humans when needed. The service has core capabilities for web quality assurance (QA) testing, data entry, data extraction, and checkout flows.

Most models today are trained in isolation, separate from the orchestrator and actuators that execute tasks, which reduces reliability. Nova Act approaches this differently by using reinforcement learning while the agents run inside custom synthetic environments (“web gyms”) that simulate real-world UIs. This vertical integration across the model, orchestrator, tools, and SDK, all trained together, unlocks higher completion rates at scale. The result is an agentic system that doesn’t merely work occasionally but is reliable at scale, with reasoning and adaptability to handle changes.

Getting started with Nova ActNova Act provides an integrated developer experience that takes you from prototype to production in hours. Let me walk you through the journey.

Start in the playgroundWe begin by visitingnova.amazon.com/actto accessNova Act Playground.There, we can quickly experiment and see Nova Act in action.

For these tests, we useNova Act Gym, a simulated browser environment designed for testing Nova Act agents. We’re using afictional travel booking websiteto terrestrial exoplanets.

Here we can quickly prototype workflows using natural language commands without writing any code. We enter the URL to automate and describe the actions Nova Act needs to perform. We can add additional actions by choosingAdd an action.

After defining the actions, we run the Nova Act agent in a live browser session. This way, we can validate that the automation approach works as expected.

After we validate the workflow, we can export it to continue development in anintegrated development environment (IDE)such asVisual Studio Code (VS Code),Kiro, andCursor.

Refine in the IDEAt this stage, we need to refine the automation in a supported IDE. We use Kiro and install theNova Actextension plugin.

The extension provides a notebook-style builder mode where we can test and debug each step individually. The live browser views show exactly what the agent is doing, while execution logs reveal the model’s reasoning and actions. This makes it straightforward to refine the workflow and handle edge cases.

To learn how to use the Nova Act extension in your IDE, visitAccelerate AI agent development with the Nova Act IDE extension in the AWS News Blog. The Nova Act extension includes templates to help you get started quickly with common workflow patterns.

With this release, the Nova Act IDE extension introduces dedicated tabs for authentication, builder mode, deployment, and running workflows—bringing the complete development lifecycle into your IDE. While the extension provides the easiest path to production, developers can also use the Nova Act command line interface (CLI) or SDK directly for more advanced deployment configurations.

Deploy to AWSWhen the workflow is ready for production, we navigate to theDeploytab to deploy directly to AWS. We enter the workflow definition name (which must match the name in the script), select theAWS Region, and optionally provide an existingAWS Identity and Access Management (IAM)roleAmazon Resource Name (ARN). The extension packages the workflow into a Docker container, pushes it toAmazon Elastic Container Registry (Amazon ECR), creates the necessary IAM roles andAmazon Simple Storage Service (Amazon S3)buckets, and deploys it toAmazon Bedrock AgentCoreRuntime.

After it’s deployed, we can monitor the workflow execution through the Nova Act console. We navigate toWorkflow definitions. The console provides observability dashbo