Amazon Nova Act

Amazon Nova understanding models deliver advanced intelligence with unmatched price-to-performance.

Amazon Nova Act is an autonomous agent model designed to interact with web browsers and applications like a human user. It can click elements, perform searches, complete workflows, and answer questions, enabling enterprises to move beyond simulation into real task execution. Delivered through an SDK and integrated with Amazon Bedrock features such as Knowledge Bases and Agents, Nova Act allows organizations to deploy agents that execute tasks reliably in production environments.

Why Nova Act?

Amazon Nova Act is built for real-world task execution, enabling enterprises to automate workflows with an AI that can act, not just respond.

Use-Cases

Organizations use Nova Act to automate repetitive business processes, test and validate web applications, execute research tasks at scale, and build intelligent copilots that complete actions across enterprise systems.

What it's best for

Nova Act is best for autonomous task automation where precision, repeatability, and direct interaction with digital systems are required.

Amazon Nova Lite - Halo Radius
Image Generated with Amazon Nova

Capbilities

Amazon Nova Act

Agent Actions

Performs browser-like interactions including clicks, form submissions, searches, and navigation.

Context

Maintains session-level awareness for multi-step tasks and workflows.

Latency

Optimized for task execution in near real time, balancing responsiveness with reliability.

Cost

Priced for enterprise-scale automation, replacing manual effort with efficient agent-driven workflows.

Your questions answered

Common questions about Amazon Nova Act

Unlike text, voice, or image models, Nova Act is an autonomous agent that executes actions such as clicking, searching, and interacting with web applications—rather than only generating outputs.

 

Nova Act is commonly deployed to automate business processes, run web-based research, validate workflows, and support copilots that interact directly with enterprise applications.

Nova Act provides AI-driven flexibility that traditional robotic process automation (RPA) tools lack, enabling it to adapt to dynamic interfaces and complex tasks without brittle, rules-based scripts.

Typical use cases include contact center automation, voice-driven copilots, interactive agents, and multilingual customer engagement platforms.

Amazon Nova Models

Compare Amazon Nova Models and Capabilities

Amazon Nova Micro

A low-latency, text-only model optimized for cost-efficient workloads. It supports over 200 languages, offers up to 128k token capacity, and can be fine-tuned for custom use cases.

Amazon Nova Lite

A high-speed multimodal model designed to handle text, image, and video inputs with exceptional responsiveness. It supports fine-tuning, processes up to 300k tokens, and works across 200+ languages.

Amazon Nova Pro

Amazon Nova Pro offers the best balance of accuracy, speed, and cost for general-purpose AI tasks. It supports 200+ languages, handles up to 300k tokens, and is fully fine-tunable for specialized needs.

Amazon Nova Premier

Nova Premier is the most advanced model in the Nova family, built for complex reasoning and high-stakes tasks. It supports 1 million tokens, over 200 languages, and is ideal for use as a teacher model in distillation workflows.

Amazon Nova Canvas

A customizable image generation model that turns text into visuals with precision and control. It supports prompts up to 1,024 characters, fine-tuning, and is optimized for English inputs.

Amazon Nova Reel

Generate high-quality video from text or image prompts, enabling fast visual storytelling. It supports up to 512 input characters and is optimized for English-language inputs.

Amazon Nova Sonic

Power real-time, human-like voice conversations with natural responsiveness and tone. It supports up to 300k tokens and is optimized for English (US, UK) and Spanish dialogue.

Amazon Nova Act

An autonomous agent model accessible via SDK, trained to interact with web browsers like a human user. It can click elements, perform searches, and answer questions—enabling real task execution, not just simulation.
Scroll to Top