Amazon Nova Act - AI Browser Automation

What is Amazon Nova Act?

Amazon Nova Act is an AWS service and SDK that lets you build and operate reliable AI agents for browser automation at scale. The agents complete repetitive UI workflows in the browser, such as form filling, data extraction, or QA testing, and escalate to a human supervisor when appropriate. You define workflows by combining natural language with Python code: an act() call passes a task in natural language to the model, while Python expresses the orchestration logic.

Nova Act addresses a concrete problem: many business processes run through web interfaces with no programmatic API or with frequently changing layouts, where traditional script-based automation breaks quickly. Instead of fragile selectors and manual maintenance, Nova Act uses a purpose-trained agentic foundation model that observes the page and derives the right actions. AWS cites over 90 percent task reliability at scale for production UI workflows. You start in the web playground at nova.amazon.com/act, develop and debug in the IDE extension, deploy to AWS, and monitor execution in the AWS Management Console.

Key Features

• Natural language plus Python: You describe tasks in natural language through act() and control the flow logic with Python code. The SDK requires Python 3.10 or higher and is English-only.
• Complete toolkit: A custom agentic foundation model, web playground (nova.amazon.com/act), Python SDK, IDE extensions for VS Code, Kiro, and Cursor, a CLI, and monitoring in the AWS Console. Workflows can be deployed to Amazon Bedrock AgentCore Runtime and Browser Tool.
• Human-in-the-loop and parallelization: Approval workflows and human UI takeover are built in. Multiple sessions run in parallel within a workflow run, so tasks scale.
• Structured extraction and tool integration: Structured data extraction via Pydantic schemas, session persistence (local files, Amazon S3, AgentCore profiles, Chromium profiles), calling external APIs and remote MCP tools, plus IAM-based credential management and S3 data controls.

Typical Use Cases

Form filling and data entry: Agents understand and populate complex web forms across multiple portals, for example updating CRM records or executing tasks in ERP systems. They adapt to varied layouts and submit entries consistently.

Search-and-extract operations: Agents navigate websites and search results, filter relevant content, and extract key fields from unstructured web sources or dashboards. This fits competitive intelligence, monitoring supplier portals, or consolidating data from platforms that lack an export interface.

Web QA and checkout flows: Agents run complete user journeys for automated UI validation and accelerate release cycles. Purchase and booking flows can also be handled autonomously from product selection through payment, for example to test retail experiences or automate recurring procurement tasks.

Benefits

• High reliability through vertical integration of the foundation model, orchestrator, and browser actuator with end-to-end training across the full stack.
• Lower maintenance effort than selector-based script automation, because the model interprets interfaces rather than relying on fixed paths.
• An end-to-end developer experience from the web playground through SDK and IDE extension to production deployment on Amazon Bedrock AgentCore with built-in monitoring.

Integration with innFactory

As an AWS Reseller, innFactory supports you with the adoption and operation of this service.