Repolode
Submit
instavm

instavm/clickclickclick

A framework for autonomous Android and computer use, letting a planner and finder LLM operate phones and desktops from natural-language tasks.

View on GitHubWebsite
Stars
714
Forks
87
Open issues
3
Last commit
7 months ago

What is it?

ClickClickClick is a Python framework that lets an LLM control an Android device or a computer to carry out tasks described in plain language, such as drafting an email in Gmail or starting a game on Lichess. It splits the work into a planner model that decides the next step and a finder model that locates the on-screen element to interact with, using ADB for Android and a macOS target for desktop.

The framework works with local models through Ollama (such as Llama 3.2-vision and qwen3.5:4b) as well as hosted models from Gemini and OpenAI, and models are configured in config/models.yaml. Tasks can be run from a CLI, a Gradio web interface, a Python API, or a REST endpoint served with uvicorn.

The project is labeled highly experimental, and the authors note the best results so far come from Gemini 3.1 Flash-Lite used as both planner and finder. Running Android tasks requires adb installed on the machine executing the code.

Key features

  • 01Uses an LLM planner to break a task into steps and a finder to locate UI elements.
  • 02Supports local models via Ollama as well as Gemini and OpenAI.
  • 03Controls Android devices through ADB and macOS desktops.
  • 04Offers a CLI, Gradio web UI, Python API and REST endpoint.
  • 05Lets you tune image quality to trade accuracy for speed.

Use cases

  • Automating repetitive phone tasks such as sending messages or opening apps.
  • Experimenting with computer-use agents using local open models.
  • Prototyping LLM-driven UI automation on Android or macOS.
  • Exposing device automation as an HTTP service for other tools.

Topics

#ai-agents#computer-use#android-automation#llm#ollama#gemini#openai#automation#agents#ai-agents-framework

FAQ

What is clickclickclick?
A framework for autonomous Android and computer use, letting a planner and finder LLM operate phones and desktops from natural-language tasks. ClickClickClick is a Python framework that lets an LLM control an Android device or a computer to carry out tasks described in plain language, such as drafting an email in Gmail or starting a game on Lichess. It splits the work into a planner model that decides the next step and a finder model that locates the on-screen element to interact with, using ADB for Android and a macOS target for desktop.
Is clickclickclick free to use? What license does it have?
clickclickclick is open source under the MIT license. Check the license terms before using it in commercial products.
Is clickclickclick still maintained?
No commits for 6+ months. The last commit to clickclickclick was on Mar 17, 2026.
What can I use clickclickclick for?
Automating repetitive phone tasks such as sending messages or opening apps. Experimenting with computer-use agents using local open models. Prototyping LLM-driven UI automation on Android or macOS. Exposing device automation as an HTTP service for other tools.
Where can I get clickclickclick?
The source code and installation instructions are on GitHub: https://github.com/instavm/clickclickclick

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects