Skip to main content

What is LiteAgent?

LiteAgent is a powerful testing framework designed to evaluate the safety and robustness of web automation agents. It provides a comprehensive suite of tools for testing how AI agents behave when encountering deceptive UI patterns (dark patterns) on websites.

Quick Start

Get LiteAgent up and running in minutes with our quickstart guide.

Key Features

Why LiteAgent?

Web agents are increasingly being used for automation tasks, but they can be vulnerable to deceptive UI patterns that manipulate their behavior. LiteAgent helps you:
  • Test Agent Robustness: Evaluate how your agents handle real-world deceptive patterns
  • Measure Success Rates: Track task completion and dark pattern susceptibility
  • Compare Agents: Benchmark different agents against the same test scenarios
  • Improve Safety: Identify and fix vulnerabilities in agent behavior

Supported Agents

LiteAgent currently supports the following web automation agents:
  • BrowserUse - AI-powered browser automation
  • DoBrowser - Chrome extension-based automation
  • MultiOn - Multi-modal web agent
  • Agent E - Enterprise web automation
  • Skyvern - Computer vision-based agent
  • WebArena - Research-focused web agent
  • VisualWebArena - Visual understanding agent
  • Human - Manual testing baseline

Getting Started

1

Clone the Repository

Clone the LiteAgent repository and initialize submodules:
git clone https://github.com/your-repo/agent-collector.git
cd agent-collector
git submodule update --init --recursive
2

Set Up Environment

Configure API keys and install dependencies:
cp collector/.env.example collector/.env
# Add your API keys to collector/.env
3

Run Your First Test

Use Docker Compose or run locally:
docker compose up --build browseruse
# or
./run.sh browseruse

Documentation Overview

Community & Support

GitHub Repository

View source code, report issues, and contribute to LiteAgent.
I