Quickstart with Notte
Create AI agents that can browse and interact with real web pages using Notte's native browser sessions, structured perception, and Gemini-based reasoning.
What You'll Build
In this quickstart, you'll create an AI agent that can:
- Launch a real browser session
- Navigate to websites and understand page content
- Take actions like clicking, typing, and scrolling
- Complete tasks described in natural language
Optional: Prepare your environment
Notte requires Python 3.11+ and uses uv
to manage dependencies and virtual environments. In a new folder, run:
# Install UV (if not already installed)
curl -Ls https://astral.sh/uv/install.sh | sh
# Add UV to PATH (macOS/Linux only)
echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.zshrc && source ~/.zshrc
# Create project folder and navigate into it
mkdir notte-test && cd notte-test
# Create a virtual environment
uv venv --python $(which python3.12)
source .venv/bin/activate
Install Required Packages
Install Notte and its dependencies, including Chromium for browser sessions.
uv pip install notte patchright
uv run patchright install --with-deps chromium
Set Up API Access
Notte uses Gemini for agent reasoning. Get your Gemini API key here: makersuite.google.com/app/apikey
Set your API key in the environment before running any script:
export GEMINI_API_KEY="your-api-key"
Create and Run Your First Agent
Create a file called agent.py
:
import notte
agi = notte.Agent(
reasoning_model="gemini/gemini-2.0-flash",
max_steps=5
)
agi.run(task="search cute cats on google images")
Run your agent:
python agent.py
You'll see the agent launch a browser, reason about the page, and take actions step by step.
What's Next?
Now that you've created your first agent, explore more advanced features:
- Try different tasks like shopping comparisons or data extraction
- Customize agent behavior with different models and parameters
- Integrate Notte agents into your applications
- Check out the full documentation to learn more