Deputy

An experimental AI coding assistant that works directly in your terminal. Deputy is a research project exploring how agentic LLM systems can integrate with development workflows, providing an assistant that can read your code, write files, execute commands, and help you navigate complex projects.

Deputy is currently experimental and built to understand how agentic LLM systems work in practice. While functional and useful, it's actively evolving toward a more robust, production-quality assistant, and at the minute is probably lacking on many fronts. Please feel free to contribute or open issues - I would love to collaborate on this

Current Capabilities

Core Features

• Interactive chat interface - Natural conversation about your code
• File system operations - Read, write, and manage files across your project
• Command execution - Run tests, build scripts, git commands, and more
• Project awareness - Understands your project structure and dependencies

AI Integration

• Multiple provider support - Designed to work with different AI providers (currently Anthropic)
• Model selection - Choose the right model for your task
• Configurable behavior - Experiment with different approaches

Installation

1. Clone and install:

   git clone https://github.com/hgrsd/deputy
   cd deputy
   cargo install --path .
   
   # alternatively, just install without cloning
   cargo install deputy
   

2. Set up your API key:

   export ANTHROPIC_API_KEY=your_api_key_here
   

   (Currently supports Anthropic - more providers planned)

Getting Started

Navigate to any project directory and start Deputy:

cd your-project
deputy

Try asking Deputy to:

• "Can you explain what this main.rs file does?"
• "Help me add error handling to this function"
• "Write a test for this module"
• "Refactor this code to be more readable"

Type exit or press Ctrl-C to quit.

Usage Options

# Basic usage
deputy

# Specify provider (currently: anthropic)
deputy --provider anthropic

# Specify model (default: claude-sonnet-4-20250514)
deputy --model claude-opus-4-20250514

# Combine options
deputy --provider anthropic --model claude-opus-4-20250514

# Use short flags
deputy -p anthropic -m claude-opus-4-20250514

# Use YOLO mode (dangerous, you know why)
deputy --yolo

# See all options
deputy --help

Current Status

As an experimental project, Deputy is actively being developed. Current areas of focus include:

• Adding support for more AI providers
• Better error handling and recovery
• Enhanced configuration options

Contributing

Contributions are welcome! Whether you're interested in adding new AI providers, improving the conversation flow, better error handling, performance optimizations, or documentation improvements, feel free to open issues or submit pull requests.

License

MIT