Features
Natural language UI automation (aiAct, aiLocate, aiAssert)
Vision-based element detection, no selectors needed
Multi-platform: web (Playwright/Puppeteer), Android, iOS, HarmonyOS, desktop
Multi-model strategy: combine planning and vision models
Bridge Mode to drive desktop Chrome directly
Chrome extension for quick experimentation
YAML script runner for workflow automation
BDD-style Gherkin script execution (Beta)
Visual replay reports and interactive playground
AI coding agent integration via Skills and CLIs
Supports open-source and custom models
AndroidWorld benchmark: 93.1% Pass@1
Caching of AI planning and location results
Consume report files for debugging
Supports Doubao-Seed-2.1, Gemini, Qwen, Kimi, Xiaomi MiMo
AI-powered code generation from natural language
One-click deployment to Replit hosting
Real-time collaborative editing
Agent 4 with faster, more versatile AI
Voice Mode for hands-free development
Custom Instructions and reusable Skills
Package Firewall to block malicious dependencies
Built-in database support (PostgreSQL, SQLite)
Integrated cloud IDE with terminal
Automatic dependency management
Version control with Git integration
Multi-language support (Python, JavaScript, TypeScript, etc.)
Context-aware code suggestions
Direct integration with Claude and Databricks
One-click MCP servers (Apollo, Black Forest Labs)
