Anthropic Computer Use (for Mac)

Anthropic Computer Use is a beta Anthropic feature which runs a Docker image with Ubuntu and controls it. This fork allows you to run it natively on macOS, providing direct system control through native macOS commands and utilities.

Caution

This comes with obvious risks. The Anthropic agent can control everything on your Mac. Please be careful. Anthropic's new Claude 3.5 Sonnet model refuses to do unsafe things like purchase items or download illegal content.

Features

• Native macOS GUI interaction (no Docker required)
• Screen capture using native macOS commands
• Keyboard and mouse control through cliclick
• Multiple LLM provider support (Anthropic, Bedrock, Vertex)
• Streamlit-based interface
• Automatic screen resolution scaling
• File system interaction and editing capabilities

Prerequisites

• macOS Sonoma 15.7 or later
• Python 3.12+
• Homebrew (for installing additional dependencies)
• cliclick (brew install cliclick) - Required for mouse and keyboard control

Setup Instructions

1. Clone the repository and navigate to it:

git clone https://github.com/deedy/mac_computer_use.git
cd mac_computer_use

2. Create and activate a virtual environment:

python3.12 -m venv venv
source venv/bin/activate

3. Run the setup script:

chmod +x setup.sh
./setup.sh

4. Install Python requirements:

pip install -r requirements.txt

Running the Demo

Set up your environment and Anthropic API key

1. In a .env file add:

API_PROVIDER=anthropic
ANTHROPIC_API_KEY=<key>
WIDTH=800
HEIGHT=600
DISPLAY_NUM=1

Set the screen dimensions (recommended: stay within XGA/WXGA resolution), and put in your key from Anthropic Console.

2. Start the Streamlit app:

streamlit run streamlit.py

The interface will be available at http://localhost:8501

Screen Size Considerations

We recommend using one of these resolutions for optimal performance:

• XGA: 1024x768 (4:3)
• WXGA: 1280x800 (16:10)
• FWXGA: 1366x768 (~16:9)

Higher resolutions will be automatically scaled down to these targets to optimize model performance. You can set the resolution using environment variables:

export WIDTH=1024
export HEIGHT=768
streamlit run streamlit.py

Important

The Beta API used in this reference implementation is subject to change. Please refer to the API release notes for the most up-to-date information.