TerminalBuddy A command-line tool that brings Apple’s Foundation Models framework directly to your terminal. Designed for macOS Sequoia and later, it enables seamless, on-device AI interactions—no app switching needed. Ideal for developers, power users, and anyone immersed in terminal-driven workflows, it puts intelligent assistance right where you work: in the command line.

TerminalBuddy Chat

• Native On-Device AI
  Leverages Apple’s Foundation Models to run powerful AI directly on your Mac—no internet connection, cloud dependency, or latency. Your data stays private and secure.

• Seamless Terminal Chat
  Interact with Apple Intelligence directly from your terminal, making it ideal for multitasking or integration with terminal-based apps.

• Streaming & Structured Output
  Choose real-time streaming responses for live feedback, or opt for complete, structured replies—ideal for scripting and automation.

• No App Switching
  Stay focused in your terminal workflow — no need to bounce between apps or interfaces..

• Flexible Interaction Modes
  Switch effortlessly between modes using intuitive CLI flags or interactive prompts, tailored to your task and preferences.

• macOS Tahoe Beta or newer
• Xcode 26.0 Beta (for building)
• Device with Apple Intelligence enabled

1. Clone the repository:

   git clone https://github.com/chaiwithcode/TerminalBuddy.git
   cd TerminalBuddy

2. Build app:

   ./build.sh

   This script compiles the tool. Once complete, the executable will be placed in the Product/TerminalBuddy folder.

3. Run the app:

    ./Products/TerminalBuddy

   execute the binary directly or install at your convience follow below method

    chmod +x install.sh
    ./install.sh 

4. Start chatting with Apple Intelligence directly from your terminal.

> terminalbuddy
🤖 TerminalBuddy — Apple Intelligence at your command
Usage:
 terminalbuddy "Your question here"
Example:
 terminalbuddy "What’s the weather in London today?"

We welcome contributions! Feel free to open issues, suggest features, or submit a pull request.

This project is licensed under the MIT License — see the LICENSE file for details.