ArkSphere Community Launched. AI-native runtime. Infra. OSS. Learn more .

blurr

An open-source Android AI operator that understands natural language and performs UI actions on behalf of the user.

Detailed Introduction

blurr (project codename Panda) is an open-source Android personal AI operator that brings LLM-driven capabilities to mobile devices. Panda can perceive the screen, understand UI elements, and perform touch actions (tap, swipe, type) via the Android Accessibility Service to accomplish multi-step tasks such as automated replies, form filling, or cross-app flows. See the project homepage and README for demos and installation instructions ( homepage ).

Main Features

  • Intelligent UI automation: combines screen parsing with natural language understanding to carry out user-level actions.
  • Local-first design: emphasizes device-side interaction and privacy, while allowing external model integrations as needed.
  • Multimodal interaction: supports voice and text I/O with configurable local or cloud-based models.
  • Community-driven: open-source repository with examples, demos, and contribution guidelines for extension.

Use Cases

  • Automating everyday phone tasks: send welcome messages, manage schedules, or auto-fill forms across apps.
  • Accessibility assistance: enable users with mobility challenges to control apps via voice or concise commands.
  • Research and prototyping: mobile agent prototype for multi-step tool use and agentic workflows.

Technical Features

  • Kotlin-based multi-agent architecture using Android Accessibility Service as the eyes-and-hands interface.
  • High-level reasoning typically relies on configurable LLM backends; README provides sample configurations and quickstart steps.
  • Gradle-built Android project with instructions for building and granting necessary accessibility permissions.
blurr
Resource Info
🌱 Open Source 🦾 Agents 🤝 Assistant 📱 Application