• English
      • New
      Midscene 1.0 is coming - now in beta

      Midscene.js

      Vision-Driven UI Automation SDK for All Platforms

      10k+
      Github Stars
      #2
      No.2 in github trending
      Get StartedDocumentation
      Platforms
      Web, iOS, Android, and more
      Control browsers and mobile apps with natural language across multiple platforms
      Unified API design for seamless cross-platform automation

      Web

      Integrate with Puppeteer or Playwright, or use Bridge Mode to control desktop browsers.

      iOS

      Use Javascript SDK with WebDriverAgent to control local iOS devices.

      Android

      Use Javascript SDK with adb to control local Android devices.

      Any Interface

      Visual modeling enables automation on any interface, beyond DOM limitations.
      MODELS
      AI Models for UI Automation
      Support for Doubao Seed, Qwen3-VL, Gemini-2.5-Pro, and UI-TARS
      Visual-language models recommended for reliable and cost-effective automation
      OpenAI SDK-compatible interface for quick integration with major model providers

      Doubao Seed

      Doubao Seed
      Vision model from ByteDance, optimized for visual understanding and UI element recognition with excellent performance.

      Qwen3-VL

      Qwen3-VL
      Alibaba Cloud Qwen vision-language model with high-quality image understanding and UI element recognition at competitive pricing.

      Gemini-2.5-Pro

      Gemini-2.5-Pro
      Google's advanced multimodal model with powerful vision capabilities and comprehensive UI automation support.
      UI-TARS
      UI-TARS
      Vision-language model specifically designed for UI automation, providing precise element localization and operation capabilities.
      DEVELOPER EXPERIENCE
      Developer APIs & Tools
      Interactive visual reports for understanding test execution
      Built-in Playground for debugging and testing
      Chrome Extension for in-browser experience

      Rich APIs

      Rich APIs
      Enables both smart automation workflows and fine-grained atomic control.

      MCP Server

      MCP Server
      Exposes device operations as an MCP Server for collaboration with various models.

      Reports & Playground

      Reports & Playground
      Enhances debugging with intuitive visualization and testing environments.

      Flexible Integration

      Flexible Integration
      Supports multiple script formats, custom models, and extensible features.
      .
      View All APIsArrow right
      aiAction, aiLocate, aiAssert...
      Explore the complete API documentation for more automation capabilities.