Blurr

Blurr is an AI agent that controls an Android device directly, performing tasks like a real person — tapping, swiping, typing, and reading the screen. Unlike traditional automation that relies on accessibility APIs or app-specific hooks, Blurr operates solely through visual analysis of the device's display, making it compatible with virtually any Android app. It's designed for developers, QA engineers, and power users who need to automate repetitive mobile tasks without modifying target applications. The system streams the device screen to a cloud or local server, where a vision-language model interprets the current state and decides the next action. Blurr handles multi-step workflows: filling forms, searching apps, scrolling feeds, extracting data, interacting with device settings. Because it mimics human interaction, it works with apps that block automated access or lack public APIs. Key features include visual screen understanding via vision-language models, human-like touch gestures (tap, swipe, long press, type), multi-step workflow automation, real-time screen streaming, and support for both real devices and emulators. It can be deployed in the cloud or on-premises, and you can customize decision-making by selecting different models. Blurr also supports data extraction from on-screen content and a headless mode for background automation. Compared to tools that rely on Android's Accessibility Service or UI Automator, Blurr is more robust against UI changes but introduces latency from screen streaming and occasional model inaccuracies. It's a powerful choice for end-to-end testing and personal assistant bots, but requires technical setup and a tolerance for imperfect reliability.