Skip to content

AppAgent

An LLM-based framework that operates smartphone apps

Visit official site

What AppAgent is

The framework is described as an LLM-based multimodal agent framework built to operate smartphone apps. It uses a simplified action space such as tapping and swiping to imitate human-like interaction without system back-end access.

What you can do with it

  • Operate smartphone apps with multimodal inputs
  • Learn new apps through autonomous exploration or human demonstrations
  • Build a knowledge base of app usage
  • Carry out complex tasks across different apps

How it works

  • A simplified action space covers tapping and swiping
  • Interactions imitate human-like behaviour on the screen
  • App operation runs without system back-end access
  • A knowledge base supports complex tasks in different apps

Application scenarios

  • Automate operations in social media apps
  • Learn how to use a new app by observation
  • Create an automated workflow across multiple apps

Pros & cons

✓ What we like

  • Apps are operated without system back-end access
  • Interactions imitate human-like behaviour on the screen

! What to watch out for

  • The described action space centres on simplified gestures
  • App usage is learned through exploration or demonstrations

FAQ

What does it operate?

The material describes smartphone apps through a simplified action space such as tapping and swiping.

How does it learn new apps?

Through autonomous exploration or by observing human demonstrations, then building a knowledge base.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review