Agent Browser

A command-line interface (CLI) tool for AI-driven browser automation. Built on top of browser-use, this tool allows you to command an AI agent to perform complex browser interactions, scraping, and testing using natural language.

Features

• Natural Language Control: Command the browser to navigate, click, fill forms, and extract information using simple text prompts.
• Multiple LLM Support: Easily toggle between powerful models like Google Gemini and Groq.
• Smart Rate Limiting: Built-in delay mechanism and fallback handling to gracefully circumvent free-tier API rate limits.
• CLI Workflows: Execute fast actions directly from your terminal.

Installation

This project requires Python 3.11+. We recommend using uv or a standard virtual environment (venv) to install dependencies.

1. Clone the repository and set up a virtual environment

   python -m venv venv
   # On Windows
   venv\Scripts\activate

2. Install dependencies

   pip install -e .

3. Environment Setup Copy .env.example to .env and fill in your API credentials:

   cp .env.example .env

   Note: Either GOOGLE_API_KEY or GROQ_API_KEY is required depending on your chosen LLM provider.

Usage

You can run the agent by giving it a specific task. To see available options, simply run:

agentbrowser run --help

Examples

Information Extraction (Fast & Simple):

agentbrowser run "Cari jadwal libur nasional Indonesia 2026 di Google dan rangkum"

Running with a Specific LLM Provider:

agentbrowser run "Cari sejarah singkat berdirinya kota Jakarta melalui Wikipedia" --llm gemini

Enabling Vision Mode (For complex sites): Note: Vision requires standard screen rendering capability and may consume more tokens.

agentbrowser run "Buka kalkulator BMI online, isi berat badan 70kg dan tinggi 175cm, bagikan hasilnya" --vision

Logs & Outputs

• Execution traces and application logs are automatically written to the logs/ directory.
• Any graphical outputs (like screenshots from Vision mode) or data exports are saved to the outputs/ directory.

Best Practices (Free-Tier APIs)

When using free accounts for LLMs (like the Gemini 15 RPM Free Tier):

• Limit your tasks to websites with simple structures (Wikipedia, News platforms, Google Search).
• Avoid massive single-page applications or heavy E-commerce sites like Tokopedia, as they can exhaust your token quota almost instantly due to their giant DOM sizes.

License

MIT License