• English
      • New
      Midscene 1.0 is released|See what's new

      Midscene.js

      Driving all platforms UI automation with vision-based model

      11k+
      Github Stars
      #2
      No.2 in Github trending
      DocumentationShowcases
      Midscene banner
      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

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

      iOS

      iOS
      Control iOS devices with WebDriver using natural language

      Android

      Android
      Control Android devices with adb using natural language

      Any Interface

      Any Interface
      Automation on any interface, beyond DOM / Accessibility limitations.
      MODEL STRATEGY

      Vision Models Multi-model combination Adapt open-source models

      Vision models boost action precision
      Multi-model setups raise completion rates
      Open-source options that still perform

      Doubao Seed

      Doubao Seed vision model optimized for visual understanding and UI element recognition with excellent performance.

      Qwen3-VL

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

      Gemini-3-Pro

      Advanced Gemini multimodal model with powerful vision capabilities and comprehensive UI automation support.
      Multi-model combo
      Multi-model combo
      Supports using different models for planning and interaction to improve task completion rates
      DEVELOPER EXPERIENCE

      Rich APIs & Tools

      Practical APIs to control automation flows and run strategy
      Supports extending your own UI action agents
      Helps developers ship UI automation tasks faster

      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
      Provides intuitive visualization reports to help developers trace back the automation process

      Flexible Integration

      Flexible Integration
      Supports using Yaml to write automation flows, supports custom Agent execution strategies
      .
      View All APIsArrow right
      aiAct, aiLocate, aiAssert...
      Explore the complete API documentation for more automation capabilities.
      DocumentationShowcases