Amazon Nova Act is a Python SDK for building and deploying highly reliable AI agents that automate browser-based workflows at scale using natural language. It allows developers to define workflows by combining natural language prompts with Python code, integrate external tools using a decorator, and manage fleets of AI agents. The library is actively developed, with the current version being 3.3.316.0, and receives regular updates.
pip install nova-actVerified import paths — ran on the pinned version, not inferred.
This quickstart demonstrates how to initialize the `NovaAct` client with an API key, navigate to a starting page, and execute a natural language instruction using `nova.act()`. The `NOVA_ACT_API_KEY` environment variable is required for authentication.
Upgrade using `pip install --upgrade nova-act`.
Be patient on the first run. For deployments, consider pre-installing Playwright browsers or setting `NOVA_ACT_SKIP_PLAYWRIGHT_INSTALL` environment variable to manage installation explicitly.
Allow the AI agent to complete its task without interference. If interaction is needed, use interactive mode or design workflows to incorporate human input explicitly.
Always use environment variables (e.g., `export NOVA_ACT_API_KEY="your_api_key"`) or AWS IAM-based authentication for managing credentials securely.
Divide large tasks into logical, granular `nova.act()` statements, each typically resulting in 3-5 browser actions (e.g., click, type, scroll).
Ensure you are in the correct virtual environment and run `pip install nova-act`.
Generate an API key from `nova.amazon.com/act` and set it in your terminal before running your script: `export NOVA_ACT_API_KEY="your_api_key"`.
Verify that your IAM policy includes the required `nova-act` permissions for the specific API action being attempted.
If encountering this with `uv`, consider using a standard Python installation or another virtual environment tool, or consult `uv` documentation for compatibility with `ensurepip`.