The first tool I am making is a simple AI browser API that works using a stealth first camoufox session and by coordinates and screenshots rather than json data, this allows an AI agent to perform multiple complex web actions rather than just fetching data, without worrying about DOM elements or complex site design.
FastAPI and Camoufox
-
Install dependencies:
pip install streamlit google-genai pydantic
-
Set your Gemini key:
export GEMINI_API_KEY="your_api_key_here"
-
Start the interface:
streamlit run gui.py