Agent.exe: Let AI control your computer directly, an open source implementation of Claude's control computer

General Introduction

Agent.exe is an open source Electron application that utilizes the Anthropic (used form a nominal expression) Claude 3.5 Sonnet API that allows users to control their local computer directly through AI. Developed by Kyle Corbitt, the project aims to provide a lightweight solution for users to experience the new features of Claude. agent.exe supports cross-platform operation, including Windows, MacOS, and Linux.

Claude especially likes the Firefox browser. If you have Firefox installed, it will perform much better.

Function List

• AI control: Through the Claude 3.5 Sonnet API, the AI can directly control the local computer.
• Cross-platform support: Compatible with Windows, MacOS and Linux systems.
• simple installation: Quick installation and configuration from the command line.
• Automated tasks: The AI can perform a variety of everyday computer tasks, such as opening a browser, managing files, and more.
• User Confirmation Mode: Provides a "semi-automatic" mode that allows the user to confirm each operation before it is performed.
• open source project: The code is publicly available and users are free to modify and extend the functionality.

Using Help

Installation steps

1. cloning project::

   git clone https://github.com/corbt/agent.exe
   cd agent.exe
   

2. Installation of dependencies::

   npm install
   

3. Configuring Environment Variables: Will .env.example Rename the file to .env, and add your Anthropic API Key.
4. launch an application::

   npm start
   

Guidelines for use

1. After launching the application, you will see a simple interface that prompts you to enter commands.
2. input commandFor example, if you open a browser, the AI will automatically perform this action.
3. User Confirmation Mode: If you enable the "Semi-Automatic" mode, a confirmation dialog box will pop up before each operation to make sure you agree before it is executed.
4. stop operation: If a problem occurs during the execution of the AI, the operation can be terminated at any time by clicking on the "Stop" button.

common problems

• The AI can't recognize the command.: Make sure your commands are formatted correctly and that the Claude 3.5 Sonnet API is working properly.
• Cross-platform issues: Although multi-platform is supported, some features may behave differently on different systems, and it is recommended to look for solutions in the official documentation.
• safety: Since AI can take full control of your computer, it is recommended to understand the risks involved and test in a secure environment before use.