Zerotap, a new Android app, aims to revolutionize the way we interact with our smartphones. The application allows an artificial intelligence to control the device, simulating the actions of a human user: tapping, scrolling, reading the screen.

Features and roadmap

The app supports various platforms and models, including Ollama, OpenRouter, Straico, OpenAI, Claude, Gemini, and DeepSeek. A chat interface has recently been added, transforming Zerotap into a full-fledged AI assistant capable of taking control of the device when needed.

The developers are looking for feedback to define the next priorities. Options under evaluation include:

  • MCP servers: to connect the chat to external tools and services.
  • Deep research: to allow the AI to browse and gather information.
  • Multi-modality: image management (reading and generation).
  • On-device models: support for Gemma 3n and Qwen, with attention to small context windows.

An interesting aspect concerns the use of Ollama: the developers ask users if they prefer to expose the instance to the internet or keep it on the local network.

The future of mobile AI

The key question posed by the developers is: what would make an AI chat on a smartphone really useful in everyday life? What are the shortcomings of current apps that support Ollama?