The Agentic Browser is an open-source AI agent designed for automating web interactions and data scraping through a natural language interface. It empowers users to automate tasks like e-commerce product searches, data extraction, and content retrieval, benefiting researchers, developers, and power users by streamlining complex web-based workflows.

How It Works

The system employs a three-agent architecture: a Planner Agent to break down tasks, a Browser Agent to execute actions via web automation tools (like Playwright), and a Critique Agent to evaluate results and guide the process. This feedback loop ensures accurate task completion by allowing agents to adapt plans and refine actions based on analysis of screenshots and DOM changes.

Quick Start & Requirements

1. Install uv: Follow instructions for macOS/Linux or Windows.
2. Clone Repository: git clone https://github.com/TheAgenticAI/TheAgenticBrowser
3. Setup Virtual Environment: uv venv --python=3.11, then source .venv/bin/activate (or Windows equivalent).
4. Install Dependencies: uv pip install -r requirements.txt
5. Install Playwright Drivers: playwright install
6. Configure Environment: Copy .env.example to .env and fill in API keys for text models (e.g., GPT-4o), screenshot analysis, Google Search, and optionally Logfire and Steel Dev.
7. Run: Execute python3 -m core.main for direct use or uvicorn core.server.api_routes:app to run as an API. Docker instructions are also provided.

Highlighted Details

• Supports intelligent web research across academic papers, travel sites, and code repositories.
• Extracts and compiles diverse data types, including sports, historical, and stock market data.
• Specializes in e-commerce data scraping (price, specifications, availability).
• Features smart, context-aware cross-domain web traversal.

Maintenance & Community

The project is built upon the PydanticAI Python Agent Framework and acknowledges Agent-E. Further community or maintenance details are not specified in the README.

Licensing & Compatibility

The repository's license is not explicitly stated in the provided README content.

Limitations & Caveats

The setup requires obtaining API keys for various AI models and services, which may incur costs. Configuration of the .env file is crucial for functionality.