On March 31, 2025, Amazon launched Nova Act, a cutting-edge AI agent designed to take control of web browsers and handle tasks autonomously. Developed by Amazon’s AGI Lab in San Francisco, this tool promises to simplify repetitive online chores—think submitting leave requests, booking dinner, or setting up out-of-office replies—all with minimal human oversight. Alongside the AI, Amazon released the Nova Act SDK, a developer toolkit that’s already making waves for its ease of use and reliability.

Unlike flashy demos with spotty success rates, Amazon is betting on dependability. Nova Act scores over 90% on internal reliability tests, outpacing rivals like OpenAI’s Operator and Anthropic’s Claude in tasks such as picking dates, navigating drop-down menus, and managing pop-ups. It also shines in benchmarks like ScreenSpot and GroundUI Web, cementing its edge in browser automation.

Available now as a research preview on nova.amazon.com, Nova Act is accessible to U.S.-based developers with an Amazon account. With just three lines of code, anyone can deploy it to automate complex workflows. Here’s what you need to know—and how to get started.

What Nova Act Can Do

Nova Act isn’t just another chatbot. It’s an AI that “drives” your browser, clicking buttons, filling forms, and extracting data as if it were you. Its core strength lies in breaking down messy tasks into bite-sized, reliable steps—called “atomic commands”—like searching, checking out, or reading on-screen info. Developers can tweak these commands with specifics, such as “skip the insurance upsell,” for pinpoint accuracy.

The AI integrates with Python, tapping into tools like Playwright for precise browser control, Pydantic for data extraction, and multi-threading for running tasks in parallel. It can even operate in the background or on a schedule—say, ordering your Tuesday salad without you lifting a finger. Amazon says it’s already powering parts of the upcoming Alexa+ upgrade, hinting at broader consumer rollout soon.

Why It Stands Out

Amazon isn’t chasing gimmicks. While competitors like OpenAI and Anthropic offer similar browser agents, Nova Act prioritizes consistency over showmanship. “We’re not after 60% success-rate demos,” the company stated, focusing instead on real-world usability. Early tests show it handles tricky web elements—pop-ups, calendars, menus—better than others, making it a practical choice for office automation or product integration.

How to Use Nova Act: A Quick Tutorial

Ready to try it? The Nova Act SDK is developer-friendly and takes minutes to set up. Here’s a step-by-step guide:

  1. Get Access
    • Visit nova.amazon.com and sign in with your Amazon account (U.S. only for now).
    • Navigate to the “Labs” section, select “Act,” and request an API key. You might join a waitlist—check back when approved.
  2. Install the SDK
    • Download the Nova Act SDK from the site or GitHub (aws/nova-act).
    • Install it via Python:
      bash
      pip install nova-act
  3. Write Your First Agent
    • Here’s a simple script to automate a task, like searching for apartments:
      python
      from nova_act import NovaAct agent = NovaAct(api_key="your_api_key_here") agent.run("Search for apartments near Redwood City train station")
    • That’s it—three lines, and Nova Act takes over your browser.
  4. Customize It
    • Add detailed instructions:
      python
      agent.run("Book a dinner reservation at Sweetgreen", instructions="Don’t add extras")
    • For recurring tasks, set a schedule:
      python
      agent.schedule("Order a salad every Tuesday at 6 PM")
  5. Run and Relax
    • Switch to headless mode for background execution:
      python
      agent.run("Submit a leave request", headless=True)
    • No need to watch—it just works.

What’s Next?

Nova Act is an early peek at Amazon’s bigger AI ambitions. The company envisions agents tackling multi-step challenges like wedding planning or IT workflows, trained via reinforcement learning for smarter, hands-off performance. For now, it’s a developer playground—available at nova.amazon.com—but its integration into Alexa+ suggests consumer tools are on the horizon.

This isn’t just tech for tech’s sake. Nova Act could redefine how we handle online grunt work, blending AI smarts with browser muscle. Whether you’re a coder automating QA tests or a business streamlining operations, it’s worth a look. Sign up, write three lines, and let it run.

By Kenneth

Leave a Reply

Your email address will not be published. Required fields are marked *