usefly 0.1.3

Agentic UX Analytics - Test your app with AI agents

Usefly

Find UX friction before your users do.

AI-powered UX testing platform. Deploy browser agents to simulate real user journeys on your web app and identify friction points, broken flows, and usability issues.

What It Does

Usefly uses AI browser agents to test your application like a real user would. Instead of writing manual test scripts, you describe what your app does and Usefly generates realistic user tasks, executes them with AI-controlled browsers, and reports back with detailed analytics on where users struggle.

Key Features

• Automated Website Analysis - AI crawls your site to understand its structure, navigation, and available features
• Smart Task Generation - Automatically generates realistic user journey tasks based on your site's capabilities
• AI Browser Agents - Executes tasks using vision-enabled AI agents that interact with your app like real users
• Friction Detection - Identifies UX issues, confusing flows, and broken functionality
• Detailed Reports - View step-by-step agent interactions, screenshots, and success/failure analysis

Installation

Choose the method that works best for you:

Option 1: Docker (Simplest)

No Python or Node.js required. Just Docker.

docker run -p 8080:8080 -v usefly-data:/app/src/data ghcr.io/dudany/usefly

Open http://localhost:8080 in your browser.

Note: The -v usefly-data:/app/src/data flag persists your data between container restarts.

Option 2: Using uv (Recommended)

uv is a fast Python package manager. Requires Python 3.12+.

# Install uv if you don't have it
curl -LsSf https://astral.sh/uv/install.sh | sh

# Create virtual environment and install (requires Python 3.12+)
uv venv --python 3.12
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
uv pip install usefly

# Install browser for AI agents
playwright install chromium

# Start the server
usefly

Option 3: Using pip

Requires Python 3.12+.

# Create virtual environment (requires Python 3.12+)
python3.12 -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install from PyPI
pip install usefly

# Install browser for AI agents
playwright install chromium

# Start the server
usefly

Open http://localhost:8080 in your browser.

Note for Contributors: If you are installing from source (e.g. cloning the repo), you must build the UI manually. See CONTRIBUTING.md.

First Run Setup

1. Navigate to Settings in the sidebar
2. Configure your AI provider:
   • Select a provider (OpenAI, Claude, Groq, or Google)
   • Enter your API key
   • Choose a model (e.g., claude-sonnet-4-20250514, claude-opus-4-20250514, claude-haiku-4-20250514)
3. Optionally adjust:
   • Max Steps - Maximum actions per task (default: 30)
   • Max Browser Workers - Parallel browser count (default: 3)

Usage

1. Create a Scenario

• Go to Scenarios → New Scenario
• Enter your target URL and a description of your application
• Click Analyze Website to let AI understand your site structure
• Generate tasks automatically or add custom ones

2. Run Tests

• Select a scenario and click Run
• Watch AI agents interact with your app in real-time
• Each agent executes assigned tasks and records every interaction

3. View Reports

• Go to Reports to see aggregated results
• Analyze friction points and failure patterns
• Review step-by-step screenshots and agent decisions
• Use Replay to watch recorded sessions

CLI Reference

usefly                    # Start server on default port 8080
usefly --port 3000        # Use custom port
usefly --reload           # Enable auto-reload for development
usefly --help             # Show all options

Supported AI Providers

Provider
OpenAI
Anthropic
Google
Groq

Troubleshooting

Common Issues

Browser agents fail to start

• Check that your API key is correctly configured in Settings
• Ensure you have sufficient API credits

UI shows blank page

• Clear browser cache and refresh

Contributing

Contributions are welcome! See CONTRIBUTING.md for development setup and guidelines.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• browser-use - The excellent browser automation library powering our agents
• LangChain - For seamless LLM integration
• shadcn/ui - Beautiful UI components